Académique Documents
Professionnel Documents
Culture Documents
by
Nagesh Sonti
DjTIC
r' 7 E
AUG 1 9 1988
SH
PENN STATE
The Pennsylvania State University
APPLIED RESEARCH LABORATORY
P. 0. Box 30
State College, PA 16804
by
August 1988
tp
. %X
i
Unclassified I 2
SECURITY CLASSIFICATION OF THIS PAGE /
REPORT DOCUMENTATION PAGE
la. REPORT SECURITY CLASSIFICATION lb. RESTRICTIVE MARKINGS
Unclassified
2a. SECURITY CLASSIFICATION AUTHORITY 3. DISTRIBUTION /AVAILABILITY OF REPORT
6& NAME OF PERFORMING ORGANIZATION 6b OFFICE SYMBOL 7a. NAME OF MONITORING ORGANIZATION
Applied Research Laboratory (If applicable) Naval Sea Systems Command
The Pennsylvania State Univers ty ARL Department of the Navy
6C ADDRESS (City, State, and ZIPCode) 7b. ADDRESS (City, State, and ZIP Code)
P. 0. Box 30
State College, PA 16804 Washington, DC 20362
E UNiGabONSRIN
ORGANIZATIONN if
1Bb. OFFICE
8. NAME OF FUNDING /SPONSORING
( EA
SYMBOL 9. PROCUREMENT INSTRUMENT IDENTIFICATION NUMBER
Naval Sea Systems Command
Sc. ADDRESS (City, State, and ZIP Code) 10. SOURCE OF FUNDING NUMBERS
Department of the Navy PROGRAM PROJECT TASK WORK UNIT
20362 ELEMENT NO. NO. NO. ACCESSION NO.
Washington, DC
17 COSATI CODES I18. SUBJ TERMS (Continue . reverse fc y and identify by block number)
FIELD GROUP SUB-GROUP aluminum alloy welding, C0 laser welding, numerical
DO FORM 1473. 84 MAR 83 APR edition may be used until exhausted. SECURITY CLASSIFICATION OF THIS PAGE
All other editions are obsolete. Uclassified
Unclassified 0
SECURITY CLASIIFICATION OP THIS PAE
coupling and to form the keyhole, than to maintain the keyhole and obtain
smooth fusion, as absorption of laser energy improves dramatically with the
keyhole formation. Laser welds in aluminum alloys contain extensive porosity,
and excessive vaporization of alloying elements occurs from the fusion zone. The
welds have severe undercuts and large discontinuous cavities on the bead surface.
A more precise control of the weld heat input and a better understanding of the
influence of process parameters and the alloying content on the weld characteristics
is required to improve the laser weldability of aluminum alloys. ,- -
aluminum alloys with marked difference in the welding process efficiencies. The
magnesium containing 5xxx series alloys showed the highest weld penetrations, as
well as periodic variation in the penetration along the weld length, evidently caused
by an unstable and oscillating keyhole. Alloy 2219 contained large shrinkage
cavities in the welds, but showed the minimum amount of voids among the
alloys investigated. Laser welds in the 5xxx series alloys had excessive amount
of spherical porosity and large irregular shaped cavities. Significant vaporization
of magnesium and aluminum from the weld, and the entrapment of parts of the
unstable cavity in the solidifying metal, are believed to cause the excessive amount
of voids observed in the 5xxx series alloys.
Unclassifipdi
SECURITY CLASSIFICATION OF THIS PA',
ABSTRACT
due to their capability of producing deep penetration welds at high welding speeds.
Although most metals are being successfully welded, laser welding of aluminum
laser energy to the material. Much higher laser intensities are required to ensure
coupling and to form the keyhole, than to maintain the keyhole and obtain
and excessive vaporization of alloying elements occurs from the fusion zone. The
welds have severe undercuts and large discontinuous cavities on the bead surface.
A more precise control of the weld heat input and a better understanding of the
influence of process parameters and the alloying content on the weld characteristics
the weld shape and dimensions, the nature and amount of porosity, loss of
elements from the fusion zone, and the microstructures in and around the
aluminum alloys with marked difference in the welding process efficiencies. The
magnesium containing 5xxx series alloys showed the highest weld penetrations, as
well az periodic variation in the penetration along the weld length, evidently caused
S'
' " " il
' b.' %'% '% % "" " " P'
. , , ... .
iv
cavities in the welds, but showed the minimum amount of voids among the
alloys investigated. Laser welds in the 5xxx series alloys had excessive amount
of magnesium and aluminum from the weld, and the entrapment of parts of the "
unstable cavity in the solidifying metal, are believed to cause the excessive amount
AceesoIn For
INTIS GFA&I
D2 TAV ~ [
'I
D, ,
Pagew
TABLE OF CONTENTS
Page
LIST OF TABLES .................................................... viii
LIST OF FIGURES.....o...................................... ix
- - ~ ~e.'.r-vVN.W'
~ .~ .'.~ \d\ .'\ , ,'
vi
TABLE OF CONTENTS (continued)
Page
Page
.5
_e
S,.N
O..
LA
55]
viii
LIST OF TABLES
Table Page
5.8 Absorptivity values used in modeling aluminum laser welds .... 147
S
ix
LIST OF FIGURES
S
Figure Page V
Figure Page
4.13 Bead shape in submerged arc weld ............................. 76
4.14 Temperature distribution on top surface after 11.5 sec (SAW) . 77
5.13 Fusion zone boundary with base metal on the right; cellular ..
structure changing to cellular dendritic in weld zone (x 200) ... 119
I
xi
Figure Page
5.16 Fusion zone boundary showing epitaxial growth from parent
metal; liquation and agglomeration of e phase in HAZ
(x 500) ....................................................... 121
5.17 Base metal in alloy 5083 (x 200) .............................. 123
5.18 Fusion zone boundary with base metal on the right (x 200) .... 123
5.19 Weld metal in the middle of the weld (x 200) .................. 124
5.20 Base metal in alloy 6061-T6 (X 500) ........................... 126
5.21 HAZ showing overaged region (x 500) ......................... 126
5.22 Fusion zone showing liquation and agglomeration
of eutectic; weld is on the right (X 500) ........................ 127
5.28 Finer equiaxed structure in the middle of weld (x 500) ......... 131
5.29 Mesh used for analyzing aluminum laser welds ................. 133 S
5.30 Laser weld pool in alloy 2219 (10 kw, 1.27 cm/sec) ............. 136
5.31 Top view of laser weld pool in alloy 2219 (10 kw, 1.27 cm/sec) 137
5.32 Long section of laser weld pool in alloy 2219 (10 kw, 1.27 cm/sec) 138
5.33 Laser weld cross-section in alloy 2219 (5 kw, 0.635 cm/sec) ..... 139
5.34 Laser weld cross-section in alloy 2219 (10 kw, 1.27 cm/sec) ..... 140
oS
xii
i:
Figure Page
5.35 Laser weld cross-section in alloy 2219 (10 kw, off-focus) ........ 141
5.36 Laser weld cross-section in alloy 5xxx (5 kw, 0.635 cm/sec) ..... 142
5.37 Laser weld cross-section in alloy 5xxx (10 kw, 1.27 cm/sec) ..... 143
5.40 Laser weld cross-section in alloy 6061 (10 kw, 1.27 cm/sec) ..... 146
9k
iv.
xiii
ACKNOWLEDGEMENTS S
giving me the opportunity to work on this project. His support, guidance and
advice during this research program have been invaluable.
for their advice and comments throughout the course of the investigation. My
thanks are also due to all the members of the Engineering Materials Department
at Applied Research Laboratory for their suggestions and help during the study.
Finally, I deeply appreciate the sacrifices that my family has so patiently endured
I. % .i
1IM
Chapter 1
INTRODUCTION
Although most metals are being successfully welded using the laser beam, :e
aluminum alloys have proved to be very difficult to laser weld.' These alloys ,'
combine high thermal diffusivity with high optical reflectivity to CO 2 laser
11 1
High intensity-.E
bpoeam 'Nkv Melted zone
(Pve) - -, ~ wi,: Wi
Molten pool
Death of weid
Pecnetatiorn (c!) I
minimal for aluminum alloys as shown in figure 1.3.' To ensure coupling, much
higher intensities are required for aluminum alloys than for steels, and at such
1.4).6 Thus, much higher laser intensities are required to ensure coupling and to
form the keyhole, than to maintain the keyhole and obtain smooth fusion.
hydrogen between the molten and the solidified metal. Porosity in arc welds
can be avoided with proper surface cleaning procedures to remove any sources of
hydrogen.8 However, such precautions have not reduced the excessive porosity
observed in aluminum laser welds. Further, laser welds in these alloys have
severe undercuts and, in some cases, large discontinuous cavities on the weld
from aluminum laser welds which in turn affects the weld microstructures and the
like magnesium and zinc can further influence the weld metal characteristics in
Due to the poor laser weldability of aluminum alloys, not much information
is available on the influence of laser welding parameters on weld bead shape and
Is
4
'.0
0.12 - 1. /,
0.04'
0.042
C0 /000 ,C
'Pi
re i n 1 the 0lc
.65
6
better understanding of the problems involved in order to improve the quality of
aluminum laser welds with improved process controls.
parameters on the final weld bead characteristics. In the thermal energy balance,
the convective terms can significantly influence the temperature profiles. However,
Oreper and Szekely1 2 have shown that for buoyancy dominated fluid flows at low
Reynold's number, the convective terms may be neglected without unduly affecting S
the weld thermal profiles. The laser welds in aluminum alloys were generally
reliably predict the weld profiles. In laser welding, there is also the phenomenon
of plasma generation at the plate surface.1 3 The plasma formation can result in
partial absorption of the laser energy. However, the influence of the plasma and
its interaction with the laser beam and the material is not yet well understood.
Its effect on the welding process is empirically incorporated in the form of overall
Surface heat sources have been effectively used for the numerical simulation of
arc welding processes. 1 4 However, these sources are not valid for deep penetration
electron beam and laser welding where the energy is deposited deep inside _0
the keyhole. In electron beam welding with high aspect ratios (ratio of weld
temperature boundariesI s ,"' have been used. Mazumder et al. 1 7 and Chande et
al.1 8 have attempted the simulation of the keyhole using the Beer-Lambert's law
using an absorptivity coefficient. Other researchers have used the weld profile to
adjust the welding heat source distribution for the modeling. For instance, Goldak
et al.1 9 ,2 0 used a model with up to eight parameters which were selected based on
7
the experimental weld geometry. Further research is required both on the physical
1.2 Objectives
Aluminum alloys are very sensitive to high energy intensity, but require such
an intense laser beam to ensure coupling and form the keyhole. The energy input
at various stages of the laser weld cycle must be optimised to improve the quality 0
of laser welds in aluminum alloys. Present investigation was therefore conducted 1'
alloys in terms of the weld penetration, top bead width, aspect ratio,
weld cross-sectional area and the process efficiency for the different welding
conditions.
tion model for simulating the melting and solidification phenomena in keyhole
laser welds.
iii. Tu predict the laser weld pool dimensions in aluminum alloys using the finite
iv. To characterize the natv-e, amount and distribution of porosity in laser welds "
N
8
v. To analyze the laser welds in aluminum alloys with regards to the vaporization
Aluminum alloys 2219, 5083, 5456 and 6061 were studied considering their
model was also developed to simulate the melting and solidification phenomena in
keyhole laser welds. Following is the layout of the thesis which describes both the
with the laser welding of aluminum alloys followed by the nature and causes of
porosity in aluminum welds. The significance of vaporization of alloying elements is
discussed with regards to the keyhole laser welding. Finally, the thermal modeling
of welds is explained with emphasis on the problems associated with the finite
and the testing procedures used in laser welding of aluminum alloys. The analytical
procedures for the selective vaporization studies in aluminum laser welds are also
presented here.
formulation, the program THERM developed, and the testing and verification of S
0
the model for stability and accuracy.
S
discussed in terms of the weld bead dimensions, the nature and amount of porosity,
loss of elements due to vaporization from the weld pool, and the metallurgical
structures in and around the weld. Numerical predictions of laser weld pool
dimensions in aluminum alloys and local solidification times are also presented
here.
6.
d.N
,.-'..
-O
10
Chapter 2
LITERATURE REVIEW
Laser welding has drawn considerable attention among the researchers with
regard to the mechanical properties of laser welds, interaction of laser beam with
the material, and the plasma plume generated at the surface of the material.
However, not much information exists regarding the laser welding of aluminum
alloys as these alloys have proved very difficult to laser weld. In this chapter,
a literature review of some aspects of laser welding of aluminum alloys and the
Factors affecting the laser weldability of aluminum alloys are discussed in the
following sections in terms of the high reflectivity to C02 laser radiation, high S
Laser irradiation of the material surface results in the heating of the substrate
surface due to the absorption of the quanta of light energy by the electrons. These
energised electrons rapidly release the energy by a process of collisions resulting "
in lattice vibrations and thereby heating. The absorption of the laser energy by
1
the surface of the material follows the relationship
a
E(x) = Eo(1 - R)e - --, (2.1)
where Eo and E(x) are the laser intensities at the surface and at depth x
the absorption coefficient. For metallic surfaces, a is of the order of 10 - ' to 10- 5
%
11 .
cm-1 , and therefore the energy absorption is limited to a very small depth of 10- 5 P
Reflectivity R of the material depends on many factors. Figure 1.2 shows the
high optical reflectivity of many metallic substances to infra-red radiation. For
material surface. Reflectivity is also a function of the surface condition and the
chemical composition of the surface. Huntington and Eager 2 2 studied the effect
found anodized and sandblasted surfaces to have improved absorptivity and the
surface. They also found that for welding applications, the joint geometry had a
23
more pronounced effect on absorptivity than the surface preparation. Jergensen
oxygen was used as the shielding gas. He related the increased penetration to
24
a decrease in reflectivity of the oxide layer formed on the surface. McLachlan
laser radiation. S
irradiated, the absorptivity improves with interaction time due to the heating and
Ready 1 found that the spectral reflectivity decrersed to about 80% of its initial
'4
12
that the increase in absorptivity does not occur at melting, but is associated
with the keyhole formation. The marked increase in the absorptivity results
from the black body characteristics of the keyhole because of multiple reflections
with progressive amounts of energy absorbed at each interaction (see figure 1.4).
25
Effective absorption of up to 90% is possible in some cases.
Further, Marcus et al."3 have reported that at higher laser intensities sufficient
to cause the breakdown of the air at the material surface, the coupling efficiency is -
consisting of electrons, ions and neutral atoms, which can partially absorb the laser '
radiation. The high temperature plasma in turn can transfer the energy to the
material. However, the nature of interactions between the C02 laser radiation and
the plasma generated at the surface of aluminum alloys is not known. Therefore,
quantitative evaluation of the energy transfer from the plasma to the material 00
laser energy due to the plasma effects can only be incorporated empirically as
Lewis and Dixon 26 , ' have shown that at sufficiently high intensities, the
plasma ignites and absorbs the incident laser energy. This ignition is associated
with the formation of a laser supported absorption wave. During laser radiation, a
._ __
13
series of laser supported combustion waves are initiated, propagated and decayed.
As the laser supported absorption wave is opaque, it can prevent the laser energy
from reaching the plate surface. It can also result in enhanced coupling by way of
are generated at intensities greater than 107 w/cm 2 , higher than the intensities
cracks, incomplete fusion, and undercuts. Most of these defects are avoided by
however, has been a persistent problem." The main cause of porosity in aluminum
welds is hydrogen picked up from the absorbed and adsorbed moisture in the
surface oxides, oil and grease layers on the surface, moisture in the shielding
gases, and the dissolved hydrogen in the base metal as well as the filler wires,
aluminum welds follows Sievert's law. 30 As seen in figure 2.1 and 2.2, the solubility
of hydrogen in aluminum at the liquidus temperature is over 19 times that at the
for steels. S
~I~F%
14
1110 1 1470
o%
O.S
0.1] - 0
A I- -
-= __I _"_ _ _
.2
S7.0
Al
Hydrogen pressure 0.01 atm
_ _
I
( j* Ni
I
WA__
30
0 loco 7.0
15G0 zcO00 I-
750 0c0
Temoeroture (*C
Figure 2.2 Hydrogen solubility in licuid metals showing
high ratio of m ximum solubi iy to melting
point solubiity for luminum.
15
However, Howden et al.3 5 have shown that in welds, the more important factor
that hydrogen is picked up over a small region of the weld pool where temperature
may be over 2000 'C. It is then distributed to the rest of the molten pool by
convection, thus causing the whole weld pool to attain a higher gas content
at the preferred site of the solid-liquid interface, but also in the molten metal
bulk liquid is possible at numerous inclusions sites available in the molten weld
metal. Uda et al. 3" also showed that the driving force for pore formation in
bubble nucleation in the liquid. Nikiforov et al."7 calculated that in the diffusive
substantially at the solid-liquid interface, and is not the main mechanism of pore
in the weld metal, and the concentration of the stable nuclei. 3 4 Porosity in welds
bubbles in the liquid metal and at the solid-liquid interface due to the rejection
of solute hydrogen, and the solidification of the liquid metal. The weld cooling
rate and hence the solidification rate and morphology significantly affect the 0
Porosity can occur in many forms in aluminum fusion welds. The rejected
gas can be nucleated both in the liquid during solidification, and in the solid
,%b...
16
immediately afterwards. 30 ,3" The former is called the primary porosity and the
latter the secondary porosity. Primary porosity in turn can manifest in two
ways. First type is the interdendritic porosity at low to intermediate levels of
occurs as small spherical pores or as long angular pores. The second type is
the large spherical pores which nucleate in highly supersaturated liquids and is
Alloying elements like Mg, Zn, Cu and Si have significant effect on the
sensitivity to pore formation in aluminum alloys. Opie et al.3 9 showed that
additions of Cu and Si reduce the solubility of hydrogen, the former being more
solubility in aluminum. Alloying elements not only affect the solid solubility, but
also the rate of absorption of hydrogen. In addition, alloying additions affect the
solidification range thereby affecting the bubble escape or entrapment.
on the nature of heat extraction. Kubo et al.4 2 have shown that simultaneous S
occurrence of shrinkage and gas evolution is required for pore formation.
In spite of the high sensitivity, weld metal porosity in aluminum alloys can
beam welding and laser welding are similar processes involving high intensity heat
sources and keyhole mode of welding, aluminum alloys are being electron beam S
17
welded, but laser welding is proving so difficult.4 3 The electron beam couples easily
to aluminum alloys, thus requiring much lower intensities as compared to the laser
beam. The high vacuum required in electron beam welding ensures the removal
of absorbed moisture and gases from the surface, and also assists in the growth
and escape of the bubbles. In keyhole laser welding, the power intensity required
to initiate the keyhole in aluminum alloys is much higher than that required to
maintain the keyhole."' Overheating of molten metal caused by the higher intensity
can have deleterious effects like increased hydrogen content in molten aluminum
and excessive vaporization of the volatile alloying elements, thereby increasing the
rotation to disperse the intense input energy that is required for coupling. The
beam rotation improved the bead surface appearance and also reduced the violent
eruptions that take place on the weld surface in keyhole laser welds in aluminum
alloys. However, beam spot rotation requires extensive modification to the
laser optics. They also found excessive penetration an. drop through in deep
penetration laser butt welds in 5xxx series aluminum alloys due to the high fluidity
of molten metal. Other researchers also report some success in keyhole laser welds
1 1 22,4 4
in aluminum alloys, but porosity was not eliminated. '
and corrosion resistance are Cu, Mn, Mg, Zn, Si, Fe, Ti, etc. Of these, Mg and
vapor pressure data for these elements. During welding, selective vaporization of
ltV %
'. P ., '.& A''' % .. +
18
the alloying elements in the aluminum alloys is expected, with significant effect on
1 investigated the
the mechanical properties of the welds. Moon and Metzbower
redistribution of the solutes and the precipitates in the fusion zone of high power
laser welds in aluminum alloy 5456. They found depletion of magnesium in the
weld, and vaporization of up to 90% of the precipitates. They also found increased
toughness in the weld resulting from the fusion zone purification, but the tensile
elements like Mn, Cr, Ni, and Al from stainless steel welds using the kinetic
theory of gases and the tabulated thermodynamic data for metallic elements and
binary alloys. They found Fe and Mn to be the dominant species which vaporized
during arc welding of these alloys. Further, using plots of evaporation power vs.
inverse temperature, they estimated the maximum weld pool surface temperature
46
to be approximately 2500 *C. Khan and Debroy used a similar analysis and
changes in the laser weld compositions in type 202 stainless steels to predict
weld pool surface temperatures. They demonstrated that the relative rates of
vaporization of any two elements from the molten pool can serve as an indicator
of weld pool temperature. Shaurer et al. used a similar approach to confirm the
47
peak temperature measurements they made in electron beam weld cavities using
an infrared radiation pyrometer. From the weld cross-sections, they measured the
cavity depth h and the base radius of the cavity r to calculate the vapor pressure
where o is the surface tension of the material, p the density, and g the acceleration
bottom of the vapor column. From the calculated vapor pressure, the associated
temperature was determined using the vapor pressure - temperature data for pure
in aluminum alloys can provide an insight into the phenomena occurring and the
work carried out in this investigation is discussed in later chapters.
To develop a predictive model for analyzing the thermal behavior of welds, the
heat conduction in the solid and the liquid metal, as well as the convective heat S
2
transfer in the molten pool must be considered. However, Oreper and Szekely'
have shown that the effect of convection in the molten pool on the temperature
profiles in the weld can be neglected if
Lz
Re- Pr k< 1. (2.3)
L
the plate thickness, Pr the Prandtl number, and k the characteristic thermal
and to a lesser extent the surface tension driven flows on the liquid surface.
3
Under these conditions, the Reynold's number can be approximated as Gr A ,
where Gr is Grashof's number and A the aspect ratio of the weld. From the
experimental results, most of the laser welds in aluminum alloys were found
to satisfy the equation 2.3, and therefore the conduction dominated weld heat
flow models neglecting the convective heat transfer terms are adequate. The
convection in the molten metal also results in rapid dissipation of the superheat. In
the Stefen problem - which involve melting and/or solidification. The problem is s
non-linear due to the moving boundary of the solid-liquid interface whose location
of the latent heat of fusion, temperature dependent thermal properties, and the
sections, some of the analytical solutions and the numerical techniques applied to
model for determining the temperature profiles surrounding a weld heat source in
steady state in a moving coordinate system attached to the heat source. Such
- '. ~% **~ ** *~ %
21
an assumption is valid over the duration of the weld cycle away from the initial
transients and end effects. He also presented a two-dimensional line source model S
applicable for through thickness welding of thin plates. Myers et al. 5 0 presented
numerous solutions based on the above model for spot welding, filler wire additions,
thin plate welding, thick plate welding, plates with finite widths, etc. Although,
point/line source models give a reasonably good prediction far from the weld,
the results near the fusion zone are far from satisfactory. Christensen et al.5 1
measured the temperature distribution in steel and aluminum welds, and compared
the results with those predicted by the point source model. They observed a
systematic deviation from the predicted results both in the weld bead dimensions
and in the temperature distribution outside the weld pool. Predictions far from
removing some of the basic assumptions. Malmuth 5 2 took into account the phase
change involved in the melting and solidification of welds. He used the matched
asymptotic expansions to solve the problem, and found that the previous empirical
method of subtracting the latent heat from the input power overpredicted the effect 6
Although, the electron beam cavity depth oscillates at a frequency of about 10 - 100
cycles/second, they reasoned that the solid-liquid interface fluctuates only slightly,
heat source. Their results agreed within 20% of the available experimental results.
Miyazaki and Geidt' 6 solved the heat conduction equation in elliptical coordinates
using an elliptically shaped heat source. Swift-Hook and Gick" used the line
ALA -1 ~n -I
22
source theory for predicting penetration in EB and laser welding. Comparison with
electron beam welds was good; however, in laser welding an order of magnitude
are necessary. Even tackling one or two of the basic assumptions results in
enormous complexity. Such solutions, therefore can only have a qualitative appeal
closer to the heat source, although some useful information can be derived for
locations far from the weld. For instance, the prediction of martensitic formation
in the heat affected zone (HAZ) surrounding the fusion welds in high carbon steels
is usually done with reasonable accuracy using the analytical models. However,
improved models are required for better understanding of the heat flow near the
weld source.
are proving to be more suitable for solving the heat transfer problems with phase
change. Time variant meshes which continuously trace the location of the solid-
For such an approach, two sets of partial differential equations with the associated .r.
boundary and initial conditions are used, one each for the solid and liquid region.
In addition, continuity of temperatures is enforced at the interface, along with a
latent heat 5
, a~T1 p
di(t) !%
kj-z - , = L d
Here, k i and k. are the thermal conductivity in the solid and liquid regions whose
Finite difference methods have been widely used for modeling the solidification
of castings, and for welding problems. However, finite element methods are
increasingly being employed for welding applications as these techniques are
readily adapted to problems with non-linearities either due to geometric or
material effects. Both methods are equally effective for simpler geometries, but 0
for complex geometries, finite element method is clearly superior. It is also easily
coupled to thermal stress problems. Gray et al. 5 compared the two methods with
regards to accuracy and efficiency. They concluded that for certain formulations,
the finite element method takes less time and is more accurate than the finite
difference method. They also found that the finite element method showed a
maximum accuracy with increasing number of nodes after which the error and Nre
computation time increased due to the round-off error in the Gaussian elimination
scheme. The computer core storage required for the finite element method is
larger. S
The finite element formulation uses either the variational method or the
weighted residual methods like the Galerkin method, least square method, or . e
the collocation method. By the usual discretisation procedure, the governing
partial differential equation and the associated boundary and initial conditions
are transformed into a set of simultaneous differential equations of the form
%
C T + K T = R.(.)
Here, T is the nodal temperature vector and "ithe vector of 11. C describes "
the heat capacity terms and includes the density, specific heat and the volume of
p.. .
24 W
the elements. K represents the conductivity matrix and includes heat conduction
between the various elements and also the convective and radiative boundary r
conditions if any. The load vector R represents the effects of the specified boundary
heat flux, specified temperature conditions, internal heat generation terms, etc.
For linear heat conduction problems, the conductivity matrix K in equation 2.2
is similar for both finite element and finite difference schemes. The heat capacity 0
Direct integration methods, rather than the mode superposition method, are
generally used in heat transfer problems in which the solution to equation 2.2 is 0
.. ,..
q~ .. # . . . - ' . - ' wm :
e",," "f " " Jr'
.0
25
obtained over a number of successive time steps. Thus, in welding heat transfer
attempted over many time steps, and due to the non-linearities involved, iterations
may also be necessary over each time step to balance the residual 'load' vector.
Although finite element discretisation over the time domain is possible,5 7 finite
difference methods are more widely used. The popular two level methods in use
.'
are the Euler's methods, the Crank-Nicholson central difference method, Wilson-0
method etc., and of the three level methods, the Gear's and Lee's methods have
been used. The choice of the technique is based on the stability characteristics,
the level of accuracy desired, and the largest time step allowed. For problems
involving melting and solidification, the above factors are intimately linked to the
apparent increase in the specific heat of the material over the solidification
range. Although valid for metallic alloys, the method requires the assumption
of an artificial but small solidification range for pure metals with a planar -
dependent specific heat for metals and alloys, and can lead to undesirable
oscillations of the solidification front. Hsiao5 8 used this method with a special 58
algorithm to avoid missing the release or absorption of latent heat and noted
iteration scheme for incorporating the latent heat directly at each time step.
ii. Enthalpy Method: Comini et al.6 have shown that the enthalpy H given by ".6 %
'N
26
the relationship
H = pc dT (2.3)
the latent heat included over the solidification range. They calculated the
volumetric specific heat from the known enthalpy at a point using an averaging
technique
dH 1 H H ay aI-a-T (2.4)
dT 3 -x aT aYaT azT
where and aSt
where 3,2 are approximated as the inverse of the temperature
material. Pham s ' has shown that this method is similar to the one proposed 0
by Hsiao. *..
time step to calculate the temperatures at the next time step, usually do not
have good stability characteristics. Implicit methods on the other hand allow .- .-.
larger time steps because of their better stability characteristics, although they
may require iterations. Myers 6 2 compared the Euler's explicit method and
W,% k ,
than the Euler's explicit method at larger time steps than the Euler's explicit
6',',
method. Bathe6 3 analyzed the two level schemes of the type V
t+AtT - tT
where
At, ...
27
ii. u = 1 is the Euler's backward or implicit method which is also first order
<
At 2
At < (1- 2a) t+aAtA"
Thus, the integration scheme is unconditionally stable provided a > 0.5, and is
only conditionally stable for a < 0.5. Wood 6 4 has described the phenomenon
severe oscillations in the Crank-Nicholson scheme and found that for fast varying
boundary conditions requiring short time accuracy, an iterative two level Galerkin
process with less severe oscillations was more accurate than the Crank-Nicholson
)S
method. Comini6 used the Lee's three level scheme with unconditional stability
in his formulation to avoid iterations. However, the formulation results in a
containing K and C. For large problems with many degrees of freedom, this ._de
inversion process can become a big constraint with the resulting loss in accuracy.
The above techniques all used the temperature formulation. The alternative
enthalpy formulation has the disadvantage that only explicit schemes can be '..
a- -p ~p e - Jk 0 .- V
28
used with the related stability problems and the shorter time steps allowed.
Pham 1 used a combination approach with the three level Lee's method. However,
his formulation precluded the consistent heat capacity approach. The Euler's
backward implicit method promises to be the most suitable for welding problems
in metallic alloys due to its unconditional stability allowing larger time steps,
of using with either the lumped or the consistent heat capacity matrix, and for
both the specific heat or the enthalpy method for the phase change problems.
The accuracy of the method is good although it has not been well documented
chapter 4.
Analytical models discussed earlier preclude the use of complex welding heat S
flux distribution, and have been limitted to point/line source or cylindrical heat
of heat flux in the welding arc. Pavelic 14 proposed a gaussian distribution model
in which the specific heat flux q(r) at any radius r about a normal circular heat
Here, qo is the peak specific heat flux which can be determined by integrating the
heat flux over the radius r from 0 to oo, and is equal to where Q is the effective
finite upper limit of integration ro where the specific heat flux q(r) drops to a
3J S
negligible value. For a value of 5%, the constant c is - -. Such a gaussian welding
r0
29
heat source can be included in the numerical model as a boundary heat flux valid
over a region of the surface, or as an internal heat source assumed present over
a small thickness at the surface. Pavelic's model has been applied to arc welding
problems with reasonable accuracy. However, the the model is clearly not valid
Goldak et al."9 presented a more complicated double ellipsoidal model for the
weld heat source in which the peak flux q is a function of x, y, z and t and is
In equation 2.7, the parameters a, b, c and f can take different values in the front
and rear part of the weld pool and Goldak chose these eight parameters based on
the experimental weld bead profile. He used the model to analyze the heat flow in
a submerged arc weld in a 10 cm thick carbon steel plate, and a deep penetration ,
electron beam weld in 1.95 cm thick steel plate. Although, the predicted weld bead
shape in the submerged arc weld was in good agreement with the experimental
results, the penetration in the electron beam weld was underpredicted by over S
40%.
energy absorption for deep penetration laser welds using the relation
e - fy
qy = 1s (2.8) 0
where is the absorptivity coefficient, and qs and qy the power intensities at the
surface and at depth y, respectively. They also assumed that once a particular -C!"
%
node reached the boiling point, it became transparent to the laser beam allowing
VP
' *"Z',%
30
the next layer to receive the energy with the energy absorption determined by
equation 2.8. The model may simulate the creation of the keyhole and thus the S
initial transient, but is not valid once the keyhole has been established and welding
A model which can more accurately simulate the deep penetration keyhole
laser welds is required and some of the attempts made in this investigation are ..
2.5 Summary
Aluminum alloys are not being laser welded due to their proneness to porosity,
poor weld bead appearance, and vaporization of alloying elements. Precise control
and alloying content on the weld characteristics , porosity formation and alloying
4oS
Chapter 3
EXPERIMENTAL DETAILS
A 15 kw AVCO CO 2 laser welding machine with F7 optics was used to weld the
aluminum alloys in the down-hand position. Autogenous butt welds were made in
2219, 5083 and 5456. Bead-on-plate welds were also made in aluminum alloy 6061
x 0.16 ") sheets in the conduction mode. All plates were in the annealed condition
prior to welding, except alloy 6061 which was in the T-6 temper. Table 3.1 gives
to welding:
i. Initial washing and scrubbing with detergent to remove dirt followed by drying
to remove moisture.
ii. Mechanical brushing using stainless steel brushes to remove oxide layers.
For butt welds, the plate edges were machined prior to above preparation.
The surfaces were then characterized by depth profile technique using Secondary
Ion Mass Spectrometry (SIMS) for levels of hydrogen, oxygen and carbon in the
cleaned and the as-received conditions of the plate. Plates used for off-focus
I
0
Zr 0.10-0.25 - - -
Others - each 0.05 max 0.05 max 0.05 max 0.05 max ,
Others - total 0.15 max 0.15 max 0.15 max 0.15 max
Al Remainder Remainder Remainder Remainder
0
3
keyhole welds and conduction mode welds were spray painted black to ensure
coupling. S
Laser power, welding travel speed and the location of laser beam focus with
respect to the plate surface were the main welding parameters investigated in the
welding experiments. Table 3.2 describes the welding parameters used for each
of the aluminum alloys. Welds were made at input power levels in the range of
5 to 14 kw. At each power level, two or three appropriate welding travel speeds
were used in the range of 0.212 to 3.387 cm/sec (5 to 80 inches per minute (ipm))
0
to obtain reasonable weld bead dimensions (ie. partial to full penetration welds).
In alloys 2219, 5083 and 5456, laser welds were made using two different focus
and (2) off-focus keyhole welds - the laser beam was focussed 0.635 cm above the
plate surface. From spot burns on acrylic plates, the beam spot diameters on the
plate surfaces for the above two cases were estimated to be 0.08 cm and 0.12 cm.
respectively. The keyhole mode welds in alloy 6061 were all made with the beam
focussed on the plate surface. For the conduction mode welds made in alloy 6061
at a separate facility, the beam cross-section obtained was rectangular (0.6 x 0.1
cm) with considerably lower power density. Further, in this setup, the depth of
focus was significantly large compared to the plate thickness, and therefore the
Helium gas was used for shielding of the molten weld pool as well as for plasma
suppression. A constant flow rate of 150 cfh was used during the welding trials. As
aluminum welds can result in drop-through due to the high fluidity of the molten
P _
34
Table 3.2. Welding parameters used.
Off-Focust 4. 10.
5. 10.
6. 10.
7.5 10.
10. 10.
"!Si,
. .. .. . . p- -, -
aluminum, all welds were made using a stainless steel backing plate to support the
molten pool.
3.3 Testing
the weld penetration, top bead width, aspect ratio, and the weld cross-sectional
area. At least three sections were examined for each weld and averaged results
are presented. The weld cross-sectional areas were determined from the macro-
photographs of the welds. The weld profiles were cut out from these pictures and
weighed. From known magnifications of the macro-photographs and the specific
weight of the paper used, the weld cross-sectional areas were calculated. -
The nature and distribution of porosity in the laser welds in aluminum alloys
of the welds. To compare the levels of porosity present in the different welds,
the volume of weld metal porosity per inch of weld length was determined by
were cut and machined from each welded plate, and were then weighed in air and
water. Comparing the results obtained with those from control specimen from the
base metal of each alloy, the amount of porosity in each specimen was calculated.
Assuming the cross-sectional area of the weld to be uniform over the specimen
length, the porosity per unit volume of the weld metal was determined. It is to
be noted that the density measurements give the total amount of voids present in
the weld metal, which may be in the form of gas pores or shrinkage cavities.
Microstructures in the weld metal, the heat affected zone (HAZ) and the parent
metal were examined using optical and scanning electron microscopes. Presence .
of any micro-porosity was also investigated. Electron microprobe analysis were
36
conducted across the weld cross-sections to evaluate the loss in alloying elements
and segregation patterns in the weld, HAZ and the parent metal.
As mentioned in section 3.3, loss of alloying elements from the laser welds was
determined for a limited number of welds using electron micro-probe analyzer. "'
was carried out. A brief description of the methodology used for developing the
P9 (3.1)
where z is the mole fraction (x i + xj = 1), and PP the vapor pressure of pure
i at the temperature of the solution. If the component i behaves ideally in the i
solution,
ai -- x i %'
,%.
where ai is the activity of the specie i in the solution. Using the tabulated vapor G
developed for each specie in the binary system of aluminum and the relevant
pressure-temperature plot for each of the aluminum alloys 2219, 5083, 5456 and
6061 can be developed. These are shown in figures 3.1(a) to 3.4(a).
IV
37
Next, the evaporation rates as a function of inverse temperature were
7
determined for each of the species using the following relation given by Dushman
where Vi is the vaporization rate in gm/cm 2 sec, M i the molecular weight in gins,
elements Cu, Mn, Mg and Zn have been included in these plots along with Al
alloy.
Next, using the method presented by Cobine and Burger,"8 the evaporation
R being the gas constant and -yi the activity coefficient (i). As the data on the
Xi0
heat of mixing as a function of temperature was not available for the systems under
M
consideration, AH was neglected in the calculation of the evaporation powers;
the error involved was determined to be not significant. Plots of evaporation power
versus the temperature are shown in figures 3.1 (c) to 3.4 (c) for the four aluminum
The above vaporization plots when applied to the power intensities involved S
in aluminum laser welds can give some important information. For instance, using
t t- 1.,
"SS
GO.
0
-~ 0
CL.
00 0 p~?
NN
LDi
00
(a -W/W QJMDNUOIDJOdA3
NN
4>
NA
Li 0 .*~.A
LIi,
US
C>,
- 0
00
L0
~)
0
C,
Lfll 0 le
.L7- l M -7 7 - -7K -
40
CN ,*
U'))
DO l 0I
po ro
- 0
100
(~wo/M)
J sM
U~JodDA~
d
__ _ _ _ _ __ _ _ _ _ __ _ _ S
41
o of
6' 0
0 0C
9- 0
C
CIAi
Q) 0
In- .
0
0-
> 4-
0L
0, 00, 0
(or*WJW6 91~ U0!'IDJodDA3
00
N"..C'.i4
-9 D"
- -U0
- 0
~ 0
CD'
oo I
o Do joD
0is~
42
the plots of evaporation rates vs. temperature as shown in figures 3.1b to 3.4b,
peak temperatures on the weld pool surface can be estimated from experimentally
a similar technique using ratios of vaporization rates of any two species in the weld
Further, the plots of evaporation power vs. temperature shown in figure 3.1c to
3.4c can be helpful in calculating the temperature profiles around the heat source
using known or assumed distribution of energy in the weld heat source, as was done
by Block-Bolten et al. 4" The dominant specie in the vapor can also be identified 'sq
at the power intensities involved. For instance, at power intensities of 3.54 x 10'
to 1.95 x 106 w/cm 2 involved in laser welding, the dominant specie vaporizing in -*5'
alloy 2219 is aluminum as seen in figure 3.1.c. However, for the other three alloys
above this transition point, aluminum vaporization decides the temperature. But,
specie, and will determine the peak temperature. The phenomenon occurring in
the weld pool are therefore intimately related to the power intensity as well as the
-
alloy content. The results obtained from the vaporization studies are discussed in
chapter 5.
,%~,
43
Chapter 4
heat transfer in the weld pool on the weld thermal profiles may be neglected for
aluminum laser welds, and heat conduction can be used to predict the weld profiles
gradients and cooling rates in and around the weld. Although finite difference -
has been chosen here to solve the weld heat transfer problem as it is more amenable
the finite element formulation, and the results of the test problems solved for
The governing partial differential equation for transient heat flow with internal
heat generation is
and z direction respectively, p the density, c the specific heat of the alloy, Q the
internal heat source term, and t the time. In addition, the following relation holds
on the surface:
k,-- q, (4.2)
anS
44
where n is the direction normal to the surface, and q' the surface heat flux. Surface
heat transfer as described by equation 4.2 can take the following three forms for
h_ = Nu-g
L9
where A', is the Nusselt number, kg the thermal conductivity of air and L the
@
the radiation heat transfer coefficient hr being
iii. Surface heat flux, eg. the welding heat source, as described in section 2.4.2.3.
0
The two-dimensional heat transfer model described below can be used to
analyze two types of welding problems. Figure 4.1 shows the geometry of the %
planar and the cross-sectional models that can be used with the corresponding
coordinate systems. The planar model is used for full penetration welds in thin
plates, and the heat flow in the thickness direction z is neglected. In such problems,
%" %
45
V
Hect Source
Hect Source
yS
SReeence lcne
w, '*1 Vm
the complete welding cycle including the initial transients and the end effects can
be modeled. However, for partial penetration welds in thick plates, the cross-
sectional model is employed in which the heat flow in the welding direction z is
neglected. This assumption is valid for high welding speeds relative to the thermal
diffusivity of the material. 9 Only the quasi-steady state conditions of the weld
cycle - away from the end effects - can be modeled by the cross-sectional approach.
Under such steady state conditions, the isotherms reach a steady state around the
welding heat source and travel along the material at the speed of welding. In other
words, following holds' 0
Although both the differential equation approach and the variational approach
can be used for the finite element formulation of the transient heat conduction
problem, the variational approach has been selected for developing this model.
This formulation allows the complicated boundary conditions to be treated p
implicitly, and only the geometric or forced boundary conditions like specified
temperatures need to be explicitly enforced.7 " The formulation used in the
model described below is essentially the same as described by Bathe.6 ' 2 '73 The .
-V! ,.r V
10
47
applicable functional fl for the variational formulation for the two-dimensional
s, being the area over which surface heat input occurs, and s2 the area over which
surface heat losses occur. The governing partial differential equation and the S
associated boundary conditions are derived by considering the stationarity of this
t K ,dV
,T' bT q,dS - bT<3 dS +j5T( PC )dV, (4.4)
where
0 ky
and
T( 8T T\
T ax ay)
Direct integration is used for the solution process whereby the time domain is
divided into several small time steps, and the solution is successively obtained at
each time step. Step-by-step incremental equilibrium equations are derived from
equation 4.4 for the solution process, where knowing the temperature distribution
in the material at time t, the new temperatures at the next time increment t + At
are calculated. Euler's backward implicit time integration scheme has been used
where the heat flow equilibrium at time step t + At is considered to solve for
temperatures at that time step. This method has been selected as it guarantees
N
48
accuracy and unconditional stability. In this time stepping scheme, the time
derivative is approximated as
t- a T- tT "
t+AtTl = t+t T(4.5)
At
dependent thermal properties and the boundary conditions involved. The equation 4.
is linearized using the following definitions
t+AtT = tT + AT,
t+AtK - tK + AK,
t+&tp tpc + Apc,
incorporated for each time step to ensure accuracy. The modified Newton-Raphson
iterative scheme has been used in which the coefficient matrices on the left hand
side of the equation are not updated for each iteration. Thus, the temperature at 'r
49
Incorporating the above linearization and the iterative scheme by substituting
r - t - dV (4.8)
region under consideration is divided into finite sized elements and the heat
isoparametric elements have been used in the model with a choice of four to eight
nodes. For any element m, the following discretisatioAl is assumed using space S
interpolation functions Hm and the gradient factions Brm as defined in table 4.1
t+AtTm H t+ArT,
mI %
t+AtT3 S 1_t+At~
t4ABTIT. (4.9)Ik
t 1t t+t
K + tc) AT% t+,ti-- t+Atci_ t+Atj (4.10)-
The individual terms involved in equation 4.10 are described in table 4.2. The
integration involved in each of these terms is carried out using the Gaussian
V.V
% giiiawlilm iiii
50
ul 0
n +
-~C -l C4
- cC 4 C4
U0
U2
I 3-
C.I C
U CY % p
C12 - I
-~ ~ _U3 vi
.a-~U2
C4 C4
I C-4 4
-~ 2C U - U.
!-Z-
~0 I
o Ri
02r~v
N.7.
%A
51
Terms Expressions
tc ~f,,H" CH dv
where
t+Atv
-v H t+Mq8 dvs
at the gauss points based on the temperatures at the previous iteration. The
set of simultaneous equations are then solved for each time step using Gauss-
Siedel method iterating till the maximum temperature increment is within a preset
tolerance limit.
Numerical welding heat transfer models using Pavelic's surface heat flux ___
model 14 as described by equation 2.6 have been reasonably successful for modeling
the arc welding processes. This model is not valid for deep penetration welds where
9
the thermal energy is deposited deep inside the material. Goldak et al. used the
double ellipsoid model described by equation 2.7 to solve the thermal flow problem
in a submerged arc weld and an electron beam weld. Although, the results for the
submerged arc weld were satisfactory, the deep penetration electron beam weld
was under-predicted by over 40%. Majumder et al. 1 7 and Chande et al.'" used a
surface gaussian heat source in their three-dimensional finite difference model for
laser welding, with the modification that once a node reached the boiling point,
it became transparent and allowed the laser energy to pass through (reflectivity
0). They assumed the energy absorption to follow the Beer-Lambert's law as
described by equation 2.8. This model may be a good attempt to simulate the
physical phenomenon of the creation of the keyhole. But, it may not accurately
represent the conditions when the established keyhole is traversed relative to the . ..
laser beam, and the energy is deposited deep inside the keyhole.
@
A variation of the above two models was used in the present study, in which
the laser weld heat input was incorporated as an internal heat source rather than *.'. '' ,
a surface flux using the following relation
2 (4.11)
q(r,y) = qo eC' e-C2Y
.
53 ,
Thus, the heat source was assumed to have a gaussian distribution in the radial
direction and with exponential decay in the thickness direction. Effective radius
and depth of the energy source, r 0 and Yo respectively, were assumed where the
intensity fell to 5% of the peak value. The various constants in the equation 2.8,
3
3Y0 I
qo = a-r-2Q
where o: is the absorptivity of the material to the laser energy, and Q the laser
power input. The internal heat source model described above was used to model
deep penetration keyhole welds in electron beam and laser welding processes, and V
TRAN language and has been developed on a Data General Eclipse MV/10000
VAX 11/782 system operating under VMS version 4.6 operating system. The
heat source either as a surface heat flux or as an internal heat source, convection
and radiation heat losses, convection in the molten metal treated as an apparent
increase in the thermal conductivity, and latent heat of fusion treated either as an
apparent increase in the specific heat of the material, or by the enthalpy method.
- ~% ni
0
54
linear or non-linear heat transfer problems. For the cross-sectional models, the
have convective or radiative boundaries. For the planar model, either of the two
faces of the elements may have convection or radiation boundaries. The same
The program allows the region being modeled to consist of up to five different
properties like thermal conductivity, volumetric specific heat and the enthalpy may
within each of these temperature regimes. The program therefore can also be used
for solving heat transfer and solidification problems in composite materials, which
An example of the program input data set giving the welding conditions
an example of the input mesh data set with explanation about the automatic
The finite element program THERM was tested for accuracy by solving several
castings, heat transfer in different welding processes like arc welding, submerged
arc welding, electron beam welding and laser welding. The results obtained from
W.~~ -0-
55
results. Following sections present the results of some of the relevant test problems
solved.
As mentioned in chapter 2, the time marching scheme most widely used for the
However, at large time steps, this method can result in stability problems. Bathe6 3
has shown that the Euler's backward implicit method results in unconditional V
stability. However, he did not present any numerical results to confirm this fact
and the accuracy of this method. Myers 6 2 has compared the finite difference and
the finite element methods using the Crank-Nicholson and the Euler's forward -.
explicit method with respect to the largest time step allowed, accuracy of the
results and the stability of the time marching scheme. He used a transient heat
To for x _Y
The same problem has been used to test the Euler's backward method used in
THERM, and the results compared with those obtained by Myers. The time
,.. .t.
After the usual discretisation procedure of equation 4.1 by either the finite
CT + K T = R. (4.13)
Incorporating the time stepping schemes shown in equation 4.11 into equation 4.12
where the matrices A and B for each of the time marching techniques being tested
Integration Scheme A B
Euler's explicit C C - K At
Euler's implicit C + K At C
Crank-Nicholson C + K _A- C - K
The vector R in equation 4.12 is a null vector as there is no nodal heat flux in,'.
this problem and the boundary conditions are homogeneous. The time step used
A t-
.. .. .. . .. . : - , -. - . . .: -..:.
::. : S
%
where k is a constant, a the thermal diffusivity assumed constant, and I the
slab. Myers has shown that the time step depends not only on the integration
scheme used, the material thermal properties and the mesh size, but also on the
The solution obtained by the finite element method using the consistent C are
shown in figure 4.2 for the Euler's implicit and the Crank-Nicholson methods. The
figure shows the temperature response at the four corners of the slab as a function
of time for k = 36. The Euler's implicit method shows excellent matching with
the exact solution whereas the Crank-Nicholson's method has a much faster time
response. This tendency is further enhanced for larger time steps represented
by smaller k of 18. For this case, the Euler's explicit method showed severe
instability, whereas the Euler's implicit as well as the Crank-Nicholson were still
temperature of T
T a for all four corners just as the exact solution does. But, the
the time response of the finite difference solutions is much slower than the exact
solution as shown in figure 4.4 even for a small time step corresponding to k -
36. The Crank-Nicholson and the Euler's implicit methods give identical solutions
for the finite difference formulation. At larger time steps, the solutions deviate
slightly from each other, although they are always stable. Only the Euler's explicit
solution became unstable at k = 4. S
S.
CC
Z! c
0 Z~
uj0- a
-00
.%
CL
I It
L. .O
0% %11 P.
-% 0
%5
% I .' 'NF
v 0,11"
T V 0*. - 1 VW ' W-I k I K
LO ~
0-'
- CL 0
0, E% -- 4
z.* IA
z u 4)
* ca 'oe
/O - - V.s
N. LO
NN
.wLr.
60
0 I,
- 0~
I I
Z C.)D C)
-i IIJ 1=
I0 0
I. 4 I .1
II II
0 %
61
Thus, the Euler's implicit method combined with the finite element method
of a molten half space, which has been used by many researchers for testing and
verifying their numerical heat transfer models. For instance, Comini8 solved the
liquid, and Hibbit 75 solved the Weiner's problem of solidification of molten steel.
THERM was tested by solving each of these problems with excellent results .
matching with the analytical solutions. Here, the comparison with the Weiner's
problem of steel solidification is presented as it models a simple but realistic
of 2845 *F is suddenly brought into contact with a half space of chill at a uniform 0
and the finite element mesh of 33 elements and 68 nodes used to model the problem
is shown in figure 4.5. The material thermal properties and the initial conditions
0
d 4P
V%e
62
* U E
0 U
LiU
zL -
0
0.- V
0
0 _ _ _ _ _ _ 0
r0
%
0 0Ile
63
Figure 4.6 shows the location of the solidification front as a function of time.
the analytical solution of Weiner as well as the finite element solution of Hibbit.
Further refinement of the mesh to 44 elements and 90 nodes did not improve the
solution indicating that the solution has converged. The lumped and the consistent
heat capacity approaches were tested using the medium mesh with no significant
with the full Newton-Raphson' scheme where the coefficient matrices are updated
Figure 4.7 shows the temperature profiles in the cast and the chill close to the S
finite element solution needs no such boundary conditions, but the temperature S
profiles show that the assumption is predicted by the model although the constant
temperature point has shifted slightly into the cast. As the solution must start
with the molten element in contact with the chill to be at same temperature as
the chill, the solidification front initiates some distance from the interface. Over-
%
prediction of the solidus at the initial stages is therefore unavoidable as seen in
tion of ingots. A cross-section of the ingot is modeled to predict the location of the
j
. r
":-:'.-:
5 " '- -
....
. ...... ., ._"-,'," :
.. . . .-.-
-- ,'.--,..- " -" "'.'--'-:.''-' - .:'Z-'', '-''. :" " - :: -'"-a *N.
64
41
LI0
ft c
.. .. .. L .-
= 0
0c
0 E
EE
C14 0
00
~0 %-
0 %
l U01000 0S 4-!osI
%%0
65 U
000 G
L. CN
cS
0 -
C) C C %
-
N,
66
solidification front as a function of time for a superheated liquid with convective
THERM was used to solve the problem with excellent matching and the results .
are presented below. 0
A 5' x 5' square quadrant of a large ingot was modeled utilizing the symmetry
in x and y direction. Two different meshes were attempted - one a uniform mesh
smaller elements near the ingot boundary. The material thermal properties and A.
initial conditions assumed are presented below:
S
Thermal Conductivity (solid) 1.0 x 10- ' btu/in sec OF;
Thermal Conductivity (liquid) : 0.9 x10 4 btu/in sec 'F;
Vol. Specific Heat (solid) : 1.0 xl0 - btu/in' 'F;
Figure 4.8 shows the location of the solidification front at times 0.4, 1.0 and -.
2.0 seconds as predicted by THERM using the non-uniform mesh, as well as the
,2., 2,
.00
67
y"
0S
-%*
-J*
: In
U Ii
AI II
- - - - - I--G
-- -- -
i ),''
/4'P 11%
e No,-
- g(N'- 0 %
% 00'f
68
solutions of Lazardis and Hsiao. The predicted results are in good agreement with
The problem of heat flow and solidification in an arc weld was next solved to
test THERM - especially the routines involving welding heat flux over a section to
7 0 has presented a finite element solution for the
of the boundary. Friedman
longitudinal butt weld in 0.254 cm x 4.0 cm plates of inconel alloy 600 with a
heat input of 703 watts and welding speed of 0.212 cm/sec. A mesh of 67 elements
and 100 nodes, shown in figure 4.9 was used for analyzing the weld; it ensured at
least six elements within the effective radius of the heat source. A uniform time
step of 0.01 sec was used from the start of the thermal analysis till the end. No
iterations were necessary within the time step before melting, and convergence was .. .-,,
achieved usually within three iterations after melting occurred. The temperature
Emissivity 0.95;
I.- "o
t "or
69
o Lfl
0 0
- %r
c
.00
%u
**
CD
C/')
0-
r- c>-
C C
< Q- V
Li, i
%C '. e
%%
* c0
,"
700
Friedman used Pavelic's disk model of surface heat flux, and assumed the .e
incorporate the heat flux on the relevant element boundaries - he applied the '.. %
welding heat as a uniform 'pressure' composed of two parts, one constant and the
other varying with time. 7 7 Further, he assumed that the heat flux was applied for
times -ia x 1.5 where v is the welding speed. In the present analysis by THERM,
the heat flux is incorporated more accurately by determining the intensities at each
gauss point of the elements concerned, and using the gaussian quadrature as shown
in table 4.2. Also, as the arc influences the reference plane only when it is r0
distance from the plane, the heat flux was applied only for the duration r
V
seconds
before and after the heat source crosses the reference plane. Thus, Friedman's 0
temperature response is expected to be spread out over a longer duration, and .. "..
have lower peak temperatures. Figure 4.10 shows the temperature response at 0.
and 0.254 cm from the weld centerline at the top and bottom surface of the plate.
The peak temperature predicted by THERM is higher by 280 K at the arc center %
As mentioned in section 4.1, the cross-sectional model allows the weld cross-
Using such a composite, the weld pool shapes at the top and bottom surfaces of
the plate was determined and are shown in figure 4.11. Here, the z-axis represents
the longitudinal axis of or the weld travel axis with the origin of the coordinate -
system attached to the center of the weld heat source. Thus z = 0 represents the
location of the arc center with the arc travelling in the negative z direction. It is
to be noted that the maximum bead width occurs slightly behind the arc center,
- ..',
- ,*.-'i', "N,. -",N
71 I
CN
CL -
00 LO
.C- Cl 2
a---0
'a v
z cc ) 4) /-
(* = = ". "
/ U)
0)
/0 0L
%%
C 00
(>, ai*ida U-
% % %
00
cop
0 E04
(Uu) t4-~ PD E 90 t0
%.
P6o
-,. ,,t
,.,,S
,34
and the maximum bead widths at the top and bottom surface do not occur in the
same plane. Figure 4.12 compares the weld profile (liquidus isotherm) predicted by
at most by 9%.
the finite element method by Goldak et al."9 using the double ellipsoid model and
by Krutz7 8 using Pavelic's disk model. The welding conditions are 38.5 kw power,
0.5 cm/sec welding speed with an efficiency of 0.95. For modeling this shallow
weld with a maximum penetration of about 1.3 cm, the surface heat flux model
was used in the THERM analysis with an effective radius r 0 of 2.0 cms. Thermal
J
0 L0
Li I
0
-- C- C-
L~J )
S II - ~ %
EO v
-oA
1Z.K
11Nb.
U %-
p oor r e
%~~ ~~ ~ ~6 ... S
rYVV1
Emissivity : 0.9;
Figure 4.13 compares the calculated weld and HAZ profiles with the exper- .N
imental results of Christensen; the predictions are in good agreement with the
experimental results. Figure 4.14 shows the temperature profiles on the top surface
of the plate 11.5 seconds after the arc has passed the reference plane. Again, the
The next problem analyzed was an electron beam weld in 1.95 cm thick carbon
steel plate, and results compared with those of Goldak et al." with the double
ellipsoid model and the experimental results of Chong"9 . The power used was 2.8
kw, the welding speed was 0.53 cm/sec, with a process efficiency of 0.95. The..-"
effective radius of the gaussian electron beam was assumed to be 0.0508 cm, and
the effective depth of the internal heat source model was assumed equal to the
plate thickness. A finite element mesh of 137 elements and 172 nodes was used
for modeling the 6.8 cm wide plate, with a finer mesh of element size 0.02 cm
near the weld source, and a progressively coarser mesh away from the weld zone.
Same temperature dependent material properties as assui.ied for the submerged "-" -
Figure 4.15 and 4.16 show the weld profiles and the heat affected zone
experimental results. It is seen that the THERM predictions match well compared
%
-t&. ;f" A
%
76
%tp
Q U,
3r C
CQ)
f, %-
,- C)
C)p < Q0 e
m M
CN)
I-I
C)%
1Ik
V Lo
w LI f
-I 00
I LO
I C.)
0 0 0
C14)
(oo) o
, %P %
,^ %- %
78
0 Go
Cl
CC)
-~~ - -0 A
0 61
w 41
% % S
AS
.
0 id
)( 0 0
C L C
UL N N N
C..,
<E <
C14,
<~ ,
C14 U ( co co
r4V
(Ujuj) UOIID I;D0.
.. .. .. ......
,'*1 , '.
80
to the experimental results with regards to the weld penetration, top bead width
NO
Locke et al. 8" have presented experimental results of high power deep
penetration laser welds in type 304L stainless steels over a wide range of power
settings. THERM was used to model one of these welds made with 8 kw laser
power and 1.27 cm/sec (30 ipm) welding speed on a 3 inch plate. The effective ,
radius of the laser beam was assumed to be equal to the beam spot radius of
0.0508 cm as reported by Locke. A finite element mesh of 136 elements and 170
nodes consisting of four and five node elements was used for modeling the 10.0 cm
wide and 1.0 cm thick plate. The material properties used are
Emissivity : 0.9;
Thermal conductivity in the molten pool was assumed to be four times that in
the solid material to approximate the convection in the molten pool. An effective
energy absorption factor of 50% was used for this weld. The weld cross-section as "
4.18. The weld penetration as well as the top bead width predicted in very good
agreement with the experimental results. The model does not exactly predict the
"+',
... +'
,,+-,,
,.,. ,. +,.r..,
-+ ,+ m '+ + ++
,,,,,+.%.
814
wuj x.
I 0LL
co
*%.
82
nail-head shape of the actual weld, but the cross-section predicted is reasonably
close.
4.6 Summary
The finite element weld heat transfer model THERM has been tested for
solved covered a wide range of input power levels as well as power intensities,
and varied from thin section full penetration welds to partial penetration welds
in thick plates. The results predicted by THERM match well with the published
experimental and numerical results. The model has been used for analyzing the 0
laser welds in aluminum alloys and the results are presented in the next chapter.
.1
83
Chapter 5
High power CO 2 laser welds were made in the four aluminum alloys over a wide
of aluminum alloys was found to be very erratic with poor reproducibility. In the
5xxx alloys, violent eruptions frequently occurred during welding, and the weld
cycle was accompanied by an intense plasma at the plate surface. In the conduction
mode welds in alloy 6061 and the lower power off-focus welds in alloys 2219, 5083
and 5456, the coupling between the laser beam and the material was sometimes
lost during welding. Therefore, all the conduction mode and the off-focus welds
were made with the plate surface spray painted black to ensure coupling. Defects
mainly observed in the laser welds were porosity, shrinkage cavities and undercuts.
Solidification cracks were occasionally found in the laser welds in 1.27 cm thick
6061-T6 plates both in the conduction and keyhole modes. This alloy is prone ON
containing Mg and Si. The problem is usually avoided using a proper filler metal
of the type 4083 with high Mg, and this aspect was not investigated.
Laser welds in aluminum alloys were characterised with regards to the weld
bead shape and dimensions, the nature and amount of porosity, alloying element
loss from the weld fusion zone, and the weld microstructures using the testing
the laser welded aluminum alloys were also simulated using the two-dimensional
finite element heat transfer program THERM described in chapter 4. The results
.4
84
5.1 Weld Bead Shape and Dimensions
The weld bead shape and dimensions described by the weld penetration, weld
top bead width, aspect ratio, weld cross-sectional area, and the overall process
efficiency are presented in the following sections in terms of the welding parameters
for each of the alloys and the laser welding modes investigated.
Weld penetrations obtained in the laser welds in aluminum alloys are shown
in figure 5.1. As expected, weld penetration increases with increasing laser power
and power intensity, and with decreasing welding speed. However, considerable
variation in the weld penetration was observed in the laser welds among the
different aluminum alloys. Alloy 2219 had the minimum weld penetration (figure
5.1a), and the 5xxx alloys the maximum (figure 5.1b and 5.1c) for similar welding
conditions. For instance, as shown in figure 5.1a and 5.1b for the on-focus laser
welds, weld penetration in alloy 2219 was 0.31 to 0.41 cm at 5 kw and 0.74 to %j,
kw and 1.14 to 1.21 cm at 10 kw. The decrease in the weld penetration with
reduction in laser power from 10 kw to 5 kw was rather pronounced for alloy 2219.
Further, the weld penetration was not uniform throughout the weld length, the
S
average standard deviation (ASD) being 0.05 cm for alloy 2219 and 0.12 cm for
the 5xxx alloys. The variation in penetration along the weld length showed some
periodicity in the 5xxx alloys as seen in the longitudinal sections shown in figure
5.10. In addition, on-focus laser welds in alloy 6061 showed considerable scatter
The conduction mode welds in alloy 6061 showed a strong dependence of weld
penetration on the plate thickness as shown in figure 5.1e. In the 0.4 cm sheets,
fJ
85
C'4.
E -
c 0
4 00
Ln - u
cli r.-
(a)O .O)J~~dP2
Ita It) t C) ut
EN
Lnn
0~
<1 0-
00
040
(wa)
PQ) or~o;,us
aM .M
n1) CD
Q.)
/N 0
0
86
Q) 1
C) -W _r
to I
0 Q)
00
(NN
U0
ai .3 q
U L O U) L) .0S
0))
000 C!
CNC
(Io 001JG
87
the weld penetration was in the range of 0.2 to 0.4 cm for power levels of 7.5 to
12.0 kw and welding speeds of 0.423 to 1.27 cm/sec (10 to 30 ipm). The weld
penetration levelled off to about 0.2 cm at the higher welding speeds. However,
for the conduction mode welds on 1.27 cm plates, the weld penetration obtained
was only 0.15 to 0.2 cm even at a higher power level of 14 kw and lower welding
The off-focus welds in alloys 2219, 5083 and 5456 were made at different
power levels using a constant welding speed of 0.423 cm/sec (10 ipm). These
welds provided some information regarding the transition from conduction mode
to keyhole mode of welding in aluminum alloys. From figure 5.1f, it is seen that
the weld penetration in alloy 2219 increased from less than 0.1 cm at 7.5 kw to
0.69 cm at 10 kw, indicating such a transition. For the 5xxx alloys, no keyhole
was formed at 4 kw for which the weld penetration was 0.13 cm. At power levels
greater than 5 kw, the weld penetration increased to almost full penetration at 10
kw.
The weld top bead widths in laser welds in aluminum alloys are shown in figure
5.2, and the data generally follows the trends observed for weld penetration. The
bead widths were found to be in the same range for alloys 2219, 5083 and 5456 at .
laser power levels of 7.5 and 10.0 kw as seen in figures 5.2a to 5.2c. However, at 5.0
kw, the bead widths in alloy 2219 are considerably smaller. The variations in bead
widths along the weld length were of the same order for all alloys (ASD of about IV
0.07 cm). The on-focus welds in alloy 6061 showed the minimum bead width as
seen in figure 5.2d. The scatter in the bead width data was also considerably less
I
TIMP Mr.c
88
CN
E -
- LO
0>0
0
f< 0
CC C 0 E
000
- 0
- c
C*
C 0
If) * I) ) It It)-0 -
ON
c 0.
00
Ec <
U 00
IVV
6w. Sw
C-i.~ C>)
-C C
W~Q CN
.4 0;
a]
C..4
E Q
(D
. 0
It)t-
CD.-o
u c
o 0 0
WN-
6 -
C.) (D 0.
/) * :3C1
LO %m
c 0.
0 o 0
<. 0 %!
C) _ Cs
% %?
90
The conduction mode welds in alloy 6061 showed decrease in bead width with
welding speed, but showed no significant dependence on the laser power. Again,
the thickness effect was very significant as seen in figure 5.2e. The bead width
data in off-focus welds was also consistent with the earlier observations regarding
the transition from conduction mode to keyhole mode as seen in figure 5.2f. Weld
top bead widths in the off-focus laser welds ranged from less than 0.16 cm in 5 kw
welds in alloy 2219 to over 1.4 cm in 10 kw welds in alloys 5xxx. The very large
bead widths in the 10 kw off-focus laser welds result from the low welding speed
From the data on weld penetration and top bead widths, the aspect ratios
of laser welds (penetration to width ratio) were compared among the alloys and
welding modes. Figure 5.3 shows the range of aspect ratios obtained in the four
aluminum alloys. Alloy 6061 showed the best ratios ranging from 0.99 to 1.36,
whereas alloy 2219 had the lowest ratios in the range 0.61 to 0.78. For alloy 5xxx,
the ratio was close to 1.0 and ranged from 0.85 to 1.27. The off-focus keyhole
welds had slightly lower ratios than the on-focus keyhole welds. The conduction
mode welds in all the four alloys had aspect ratio of about 0.4.
4 to 5. Aluminum laser welds resemble more the plasma arc welds in steels. Alloy
6061 with the smallest top bead widths for similar welding conditions resulted in
the best aspect ratio, whereas alloy 2219 with the smallest weld penetration also
I
91
=E 2
LO 0
EE
000
V)
C14)
ll'S
oijo8loods
92
5.1.4 Weld Cross-sectional Areas
Although weld penetration and bead width data are important in selecting the ,
weld parameters and for design considerations, the influence of welding parameters
on the weld cross-sectional area gives more information on the process efficiency
and absorption characteristics of the material. Figure 5.4 shows that the data
on weld area for the different aluminum alloys and welding modes is consistent
with trends observed earlier in the penetration and bead width data. Weld area
increases with laser power and decreases with welding speed. Alloys 2219 and 6061
had the minimum weld area, and the 5xxx alloys the maximum for the keyhole
welds. For instance, in 7.5 and 10 kw on-focus keyhole welds in alloy 2219, the
weld area was about 58.8 to 69.6% of weld areas in alloy 5456; at 5 kw it was even 0
lower (12.6 to 21.8%). Weld areas measured for off-focus welds in alloys 2219 and ,
5xxx also confirmed the transition from small conduction mode welds to larger
The overall welding process efficiency for the aluminum laser welds was
determined by calculating the quantity of heat (ho) required to heat and melt
the volume of weld metal per cm of weld length. The ratio of h0 to the total laser
heat input h i per cm of weld length (h- = P--e) is defined here as the process
efficiency. Thus, the process efficiency is given by the expression
v(cAT + H)
power/speed S
where v is the volume of the weld metal per unit length of weld, c the average .
volumetric specific heat, AT the difference between the ambient temperature and
the melting point, and H the latent heat. 'U'
,o
'U-.):
93 d9 QJ
It'
I.) U -'
0 A
000
I n 0
m0
0 - E
Loo LO LO E
C) U
00
bsN 00opa
I 0
0 .1-
94
.04 C, S
C14 U' __
CN W
-k LOnq
CVV
0 0 01: 0
< u
.1.
kn * n U") It) Q)
* N -CNJ
''3 ~. *
__O -O Ul
.~I -r r T, ~
Figure 5.5 shows the plots of process efficiency in laser welds in aluminum
alloys. In on-focus laser welds, process efficiency increased with laser power, but
showed no consistent trend with respect to the welding speed. Keyhole laser
welds in alloy 5456 showed the highest process efficiency of 30 - 33% at 10 kw,
and between 18 and 24% at the lower power levels. Alloy 5083 showed slightly
lower process efficiencies, but with similar trends as alloy 5456. For alloy 2219,
process efficiency decreased drastically from 20 - 23% at 10 kw to only about 2 -
focus welds in alloy 6061, the process efficiency ranged from 17 - 25% at all power
levels investigated, whereas in the conduction mode welds in the same alloy, the
process efficiency was always less than 5%, and generally decreased with welding
speed. Off-focus welds showed a wide range of process efficiencies from as low as
0.6% in conduction mode welds in alloy 2219 to over 20% in keyhole welds in alloy
5083.
The process efficiency plots in figure 5.5 show some inconsistencies. For
instance, 77 decreases with welding speed at 5 kw, but at 7.5 kw, 7 increases
with welding speed. At 10 kw, a maximum q is observed at 1.27 cm/sec (30 0
ipm). The calculation of 7 involves the volume-of weld metal v per unit length of
weld, which is dependent on the welding speed and the laser power as discussed in
section 5.1.4. However, the role of the welding parameters on the weld volume is
rather complex. At low welding speeds, a larger amount of heat loss occurs due to
conduction with resulting low 77. Similarly, at high welding speeds, the 17 is lower
as there is insufficient time for the material to react. The weld metal volume v is-
the laser power. This manifests as the apparently inconsistent behavior observed
96
LO
-4) 0
0 1
c0
O~~ C ) LO O
CAOUIOI;J3 SGOOJd :
U C
0i4 .0 -
W LQ 0
0 U)
E' 0
< 00
in 0 LO) 0 fl 0 l c
PI N (N (D-
C)AOU'O';Jj3 SS9QOd pI
-4
U _ 3
E o
CL
Ln
0
.) <
pi C .N
I-
97
CNN
LO In It
c 0
I -d C)J
(~) Au~io~j3
S~0.J
09
U-0 CD O a -O
C-4
A (I)
00
< u
'0
C: 13
040
CL :
0.E)
0
0v A.0
0 -0
C 1)J)b ~ fl
ft'~ - ~ ~ 4
In CN C
A~uQOl~i SSDO~d!Z9
.ftr
98
in figure 5.5. A schematic of the expected behavior of process efficiency is shown
in figure 5.6.
5.1.6 Summary
Aluminum alloy 2219 showed considerably lower weld penetration than the
5xxx series alloys or the alloy 6061. As mentioned in section 2.1.3, the balance
of forces at the bottom of the keyhole given by equation 2.2 relates the vapor
pressure P, with the cavity depth h. Due to the presence of volatile elements
like magnesium, alloys 5xxx are expected to result in higher vapor pressure at
the bottom of the cavity than in alloy 2219 for similar welding condition (see
figures 3.1 to 3.3). Thus deeper cavities are expected with consequently higher
Generally, the laser welds in aluminum alloys exhibited much larger bead
these alloys is a key factor in causing such large bead widths. Further, the energy
required to melt a unit volume of weld metal in aluminum is about 30% of that
required for steels. The laser welds generally depicted a nail head shape with
large bead width at the top surface which reduced rapidly to less than half the top
width at depths slightly below the plate surface (see figure 5.9). One of the causes
of such a weld shape is believed to be the reradiation of energy from the plasma .-
generated at the plate surface. In addition, the laser optics also influences the
weld profile obtained. The focal lengths resulting from electromagnetic focusing
in EB welding are much longer than in laser welding. The shorter focal lengths
involved in laser systems in turn result in larger beam divergence angle, and the
r
-4-
CN Q)
C:
Q-
_00
u-
0
Q))
0)0
*00
0))
u)
L-
SSQOO0)
100
Weld bead dimensions described in the last few sections showed that the laser
welds in aluminum alloys ranged from conduction mode to keyhole welds. The
overall welding process efficiency data was used to classify the welding mode for
each of the welds. The on-focus welds in alloy 5xxx and 6061 had process efficiency
better than 17%. However, alloy 2219 exhibited a process efficiency of only 5% or
less for the 5 kw on-focus welds as compared to over 13% efficiency at 7.5 kw and
over. Absorptivity of these aluminum alloys is in the range of 5%, and it improves
considerably once the keyhole is formed. The weld bead dimensions as well as
process efficiency data in off-focus welds in alloys 2219 and 5xxx also showed a
transition from conduction mode to keyhole mode. Thus, a process efficiency value
of about 5% was defined as a threshold below which the welds were classified as
conduction mode welds and above which the welds were classified as keyhole welds.
Using such a criteria, the threshold power intensity for keyholing for alloy 2219 was
estimated to be 6.63 x 105 to 8.84 x 105 w/cm 2 . Similarly, the threshold intensity
for alloys 5xxx is in the range of 5.3 x 105 to 6.63 x 10s w/cm2. In conduction
mode welds in alloy 6061, no keyhole was formed even at 14 kw, and therefore,
2
the threshold intensity is greater than 2.33 x 105 w/cm .
torr. A typical depth profile for aluminum alloy 5083 for both the as-received and
101 I
cleaned surfaces is shown in figure 5.7. The x-axis represents the sputtering time
S
in minutes, and the y-axis represents the detected intensity in counts per second
(c/sec) of the different specie being investigated. To determine the actual hydrogen
content present, standard specimen of each alloy with known levels of hydrogen are
required. Further, the depth of sputtering may not be uniform among the alloys
as the sputtering rate varies with the alloying content. As standard specimen were
not available for the four aluminum alloys, the relative intensity of hydrogen with
respect to aluminum was compared for the two surface conditions for each alloy
The intensity of hydrogen (+) has a maximum (peak) value at the outermost
surface, and it drops to a lower (base) value within a few angstroms in the material
as shown in figure 5.7. The peak and the base values of the relative hydrogen
intensities are compared for the as-received and cleaned surfaces for each of the
alloys in figure 5.8. The surface preparation procedure has significantly reduced
the hydrogen intensity at the outer surface. Further, although the peak and base
values of hydrogen intensities vary considerably among the four alloys in the as
received condition, the base levels are quite uniform among these alloys in the
cleaned condition, thereby allowing the volume of porosity to be compared for the
1%'
Figure 5.9 shows transverse sections of typical laser welds obtained in the
occasional shrinkage cavities were observed which were sometimes as large as 0.3
to 0.4 cm in diameter and usually located near the middle of the weld. Otherwise,
U.':l_.
%iS ;
102
Is
I-G
to
as 3 s it* 3 .,
a~ 2
AIM2
1031
0)
C.)4
co 2
L) 0
1)
0 C)0 %t
too
2 L-
a)/3 BIS0u
F.R pu- K,
104
a. Alloy 6061
10 kw
1.27 cm/sec c .
(30 i pm) t
b. Alloy 5456
10 kw
0.843 cm/sec
(20 ipm)
C. Alloy 5083
10 kw
0.843 cm/sec
(20 ipm)
10 kw
0.843 cm/sec-
(20 ipm)
105
these alloys did not exhibit any significant porosity under microscopic examination
S
of transverse sections even at high magnifications. However, the keyhole welds in
Mg containing alloys 5xxx contained large amounts of spherical porosity with pore
diameters ranging from 0.01 cm to over 0.3 cm. The larger spherical pores were
usually clustered near the top surface of the weld bead, and the smaller pores
distributed throughout the weld. In addition, these welds also contained frequent
large cavities trapped in the middle of the weld. The off-focus keyhole welds in
alloys 2219, 5083 and 5456 also showed similar results to on-focus keyhole welds.
However, the off-focus conduction welds in alloy 2219 as well as the conduction
mode welds in alloy 6061 were generally free of any kind of porosity or shrinkage
cavities. The 5xxx series alloys exhibited distributed small pores of about 0.01 cm
Longitudinal sections of the welds shown in figure 5.10 reveal the worm-hole
nature of the porosity, as well as the severity of porosity in the 5xxx series alloys. %
of the alloys. The volume of porosity determined from the density measurements
accurate to within 0.5%. Alloy 2219 contained the minimum amount of voids
present, whereas alloys 5xxx had the maximum. Following paragraphs briefly
describe the observations regarding the porosity data for the different alloys.
As mentioned above, this alloy showed the least amount of porosity, in the %
range of 0.019 to 0.067 cc per linear inch of weld. Figure 5.11a shows that at
laser power levels of 5.0 and 7.5 kw, porosity increased with welding speed, and at
(30 ipnl) S-
b.>A 08 Ally
10 kw-
02
-
1.2
Aloy503 '
(30 i pm)
c. AllIoy 5456 z
10~ kw .
1.27 cm/sec
(30 i pm)
C4
V)
0 ~%
c ~0 U
-~ coc
-
-n --
00
0 0 0
(~~~~)C
O~uno~,- Acoo
<o
0\**
C4
~'
atO r, C)a
-z
0
E -
(Io
w 0
(0)A~ISO.JOd ;o awnOA
108
-ft
E
C)
'4 D- ,
00
< 0)
0
; o~. o w n l. ojCO~
(~) ~'i~~o~
O 9Wl~Ou
CL)
PI)~
1))
-- 0
V) (D
0 A.t.
0 0~
01 -
0 LO
(A-~ISOjOd ;o awrngOA
U-% %
%,r.
4."
. %
.%
109
10.0 kw the trend was opposite. Moreover, porosity at 5.0 kw was the minimum,
whereas at the higher power levels investigated, porosity was about equal. The
variation in porosity along the weld length was very small, the average standard
Alloys 5083 and 5456 had higher levels of porosity in the range of 0.09 - 0.36
cc as seen in figures 5.11b and 5.11c. Porosity decreased with increasing welding
speed at all power levels unlike in alloy 2219. Also, maximum porosity was found
at the lowest power level of 5.0 kw. The variation along the weld length was also
Porosity was observed only in the keyhole welds, and was in the range of 0.08
to 0.16 cc (see figure 5.1ld - note that the welding conditions are slightly different
from the earlier welds). In general, porosity decreased with increasing welding
In the off-focus kevhole welds made at a welding speed of 0.423 cm/sec (10 B
ipm), porosity increased with increasing power for all the alloys as seen in figure
5.11e. Only in the 5xxx non-keyhole welds, a small amount of micro-porosity was
5.3.5 Summary
Conduction mode welds in alloy 6061 as well as the low power off-focus welds
in the other three alloys were generally free from porosity and large cavities. The
r
110
pool was found to be effective by the depth profile techniques using SIMS. It is
reasonable, therefore, to assume that the porosity and voids observed in aluminum
laser welds are not caused by hydrogen. Porosity and cavities were observed only
in the keyhole mode welds at higher laser intensities. These keyhole welds are
dominated by the vaporization of elements from the weld pool and the formation
of vapor column.
In 5xxx series alloys, the weld penetrations were found to have cyclic variations
with a frequency of about 4 to 8 hz as seen in figures 5.10b and 5.10c. This variation
in these alloys. For a partial penetration weld, the condition for stability at the
PV = 2 + pg h,
2
aid at a depth of x from the plate surface by the expression
-=-+pgz.
r
Assuming that the vapor pressure increases linearly from 0 at the plate surface to
P, at the bottom of the cavity, the condition for a stable keyhole is given by a 7-A
maximum depth of cavity -A
"-...
.' .. ' -,.
hmaz r x Pv(h)
(a + p r g x)
If the cavity depth is larger than hmaz, the cavity collapses with th liquid metal
closing the cavity at the depth x. The laser beam then impinges the liquid
metal and the cycle of keyhole formation and collapse repeats at the characteristic
frequency.
Irb
Iil
of cavities present in the laser welds in alloys 5xxx. In the alloy 2219, there
is no evidence of an unstable keyhole, and the weld penetration does not vary
significantly over the length of the weld as seen in figure 5.10a. This alloy
consequently contained much smaller amounts of voids present in the form of "
shrinkage cavities.
The vaporization of alloying elements in the weld metal due to laser welding
0
was investigated for a few selected welds using the electron micro-probe analyzer
(EMPA). To compare the gross change in composition in the weld metal due to
welding, numerous random spot measurements were made in the base metal as
well as the weld metal. Due to the microstructural heterogeneity in the aluminum
alloys, use of small beam spot of one micron diameter usually gave erroneous
results. Thus, the random spot measurements were made using a larger beam
spot of 20 microns. At least ten measurements were taken for each weld to ensure
accuracy. The error involved in the EMPA measurements is usually of the order
The most significant change in the composition (weight percentage) in the weld
zone in alloy 2219 was in the two major species aluminum and copper. Table 5.1
shows the chemical composition of the base metal and the on-focus laser welds.
@
112
Weld A] Cu Fe Mn Si Ti
Depletion of aluminum from the weld metal in alloy 2219 is expected. Vapor
the melting point to boiling point range of aluminum and beyond as shown
in figure 3.a. Also, from figure 3.1c, for the range of input power intensities
involved, aluminum boiling determines the weld pool peak temperatures, and is
copper from the molten pool, copper enrichment observed is the result of loss of
aluminum. Based on the above assumption, the percentage loss of aluminum was
%0
113 ~
combined with the volume shrinkage of the molten pool on cooling can cause the I
Table 5.2 compares the percentage loss of aluminum with the volume of
porosity determined by the density measurements for a few selected welds and
shows a direct relation between these two effects. Although welds 102 and 107 have P rP
been made at different welding conditions, they still result in the same amount
of heat input per unit length of the the weld. However, they show considerable
present. The higher peak temperatures in the welds associated with the higher
input power intensity involved in weld 102 has resulted in the larger amount of
aluminum vaporization as compared to the weld 107. Similarly, weld 103 made at
to weld 102 made at 10 kw and a lower welding speed of 30 ipm. The higher
welding speed results in lower power input per unit length of weld, with generally
In the 5xxx series aluminum alloys, the most significant change in the weld ,
metal composition was observed in aluminum and magnesium which were both
40
5.3 and 5.4 summarize the results of compositional changes in a few on-focus
keyhole welds in alloys 5083 and 5456 respectively. Again, using similar arguments
as in section 5.4.1 and from figures 3.2a and 3.3a, it is seen that manganese has
the melting point to boiling point range and beyond. It is therefore reasonable
element is solely caused by the loss in aluminum and magnesium. Based on this
assumption, the loss in aluminum was estimated to be from 7 to 19% in alloy 5083,
and up to 22% in alloy 5456. Similarly, the loss in magnesium was found to be
Weld Al Mg Mn
93.23
4.29
4.26
1 0.77
0.66
.., .
(# 204) "__
%
i
115
Weld Al Mg Mn
Table 5.5 and 5.6 compare the percentage loss of aluminum and magnesium
as well as the volume of porosity observed among the different welds in alloys
5083 and 5456 respectively. Several inconsistencies are evident among these alloys
as well as within each alloy. For instance, comparing welds 201 and 203 with
increasing welding speeds for the same input power of 10 kw, percentage loss in
aluminum increased from 12.49 to 19.19, but the percentage loss in magnesium
decreased from 26.62 to 22.25, and so did the porosity volume. However, between
welds 201 and 207 with equal power input per unit lengths but different welding
conditions, both the percentage loss in aluminum and magnesium decreased, but 0
the porosity volume increased. Further, comparing the welds 201 and 203 in alloy
5083 with the welds 302 and 301 in alloy 5456, although the trend in porosity A.
volume is similar, the trends in percentage loss of aluminum and magnesium are
reversed.
It is clear that the higher magnesium content in the 5xxx series alloys results
in much larger amounts of porosity in the weld as compared to the magnesium free
alloy 2219. But, the porosity volume could not be correlated to the percentage loss
" . . . . . . . . . . . . . . . . ..-.. .1 .
._J 116
Table 5.5. Loss of aluminum and magnesium related to porosity in alloy 5083.
Table 5.6. Loss of aluminum and magnesium related to porosity in alloy 5456.
5.4.3 Summary _
Porosity and shrinkage cavities were observed only in the higher intensity
well as other volatile alloying elements. A direct corelation was observed between
the loss of aluminum and the volume of porosity measured in alloy 2219. This alloy
exhibited large shrinkage cavities in the middle of the weld and very little amount N
of spherical gas porosity of the kind observed in the 5xxx alloy. The 5xxx alloys ",N
showed a large amount of spherical porosity with large pores accumulated near the
top surface of the welds. The large bubble formation is attributed to the boiling
-p
m
117
considerable amount of the bubbles escaped from the molten metal, but sufficient
S
quantity of bubbles were trapped in the weld resulting in the large volume of
spherical porosity observed in these alloys in addition to the large cavities in
the middle of the welds. The aluminum alloys have a large solidification range,
which results in a large mushy zone during solidification as seen in figure 5.35 and
5.36 which are discussed later. As solidification progresses, the large mushy zone
prevents the bubbles to float up and escape, resulting in the entrapment of the
large amount of spherical pores.
The weld metal properties are dependent upon the microstructures and
segregation patterns in the weld zone, and are a function of the welding process
variables, the weld geometry, and the material thermal properties. The welding
process also effects the properties of the base metal surrounding the weld zone
called the heat affected zone (HAZ). In the non-heat treatable aluminum alloys in
the series lxxx, 3xxx, 4xxx and 5xxx, the strengthening mechanism employed is
cold working. Thus, for these alloys, the welding cycle effects the HAZ by annealing
it to varying degrees of strength reduction. However, in the heat-treatable
aluminum alloys belonging to the series 2xxx, 4xxx, 6xxx and 7xxx, the principal
S
strengthening mechanism is the controlled precipitation of the supersaturated R%
solutes in the alloy. This is achieved by solution heat treatment and quenching
to dissolve the solute into the matrix, followed by an aging process to induce.,.- .
therefore results in a variety of structures in the HAZ ranging from the fusion
zone, solid solution zone, overaged zone to unaffected base metal. In the following
paragraphs, typical structures observed in transverse sections of a few selected
welds are described.
da
9
....
118
figure 5.12. The microstructure consists of elongated grains in the rolling direction
of the solid solution with small particles of CuA12 (light) widely distributed,
The weld zone for a 10 kw, 20 ipm on-focus weld is shown in figures 5.13
network of the E phase (CuA12 ). However, the substructure varies widely from
the edge of the fusion zone to the center of the weld. In figure 5.13, the edge of
the fusion zone is shown with the unmelted region in the bottom right side of the
figure. The weld zone consists of a small region of cellular grains changing rapidly
structure in more detail and clarity. Finally, figure 5.15 shows equiaxed grains
from the interior of the weld far from the weld boundary. The primary dendrite
arm spacing in this equiaxed region is finer than in the outer edges of the weld, %
the dendrite arm spacing does vary considerably within this equiaxed region. It
is to be noted that for a reliable measurement of the dendrite arm spacing, the
the outer regions of the welds where columnar dendrites are observed with growth
direction clearly defined. However, in the interior of the weld with the equiaxed
grains randomly oriented, average values of dendrite arm spacing were taken from
many grains.
0v
luxu
119i
.
+ - -... -l.
,. . .... g
- is=4j
jS..
S * - 7 . .-.
V . . ..
,. .,._ -. ,, =,..
Figure 5.12
-
.,
Bose metal
- --. . ...
2219 (x 100).
in alloy "" - ' V:,
F ."
w
'%
..
Alp %~ % P
2* %
120
I.IA\
~~
-'
-4,
A-
columnar ~ i i ~ st u t .ed
r (x 2 0)
.%.% %
Uv p rS
A A A -
- - - - . *~ 4 t-%. A
121
16 Fsio shwin
zoe 5bonday
Figue ep taxil gowt
of
HA 00)
(xpase n
I -.
_k .... 4
122
In the HAZ close to the fusion boundary, there is evidence of incipient melting
as well as agglomeration of the 0 phase as seen in figure 5.16. As this alloy was
originally in the annealed condition, other structures like the overaged zone are
not present.
From figure 5.9a showing the macro-section of the weld, the weld appears to
regions interspersed with equiaxed grains are evident. Such banding was unique
to this alloy and was observed in both the on-focus as well as off-focus welds.
As the composi.,ons of the aluminum alloys 5083 and 5456 vary only
marginally, the weld microstructures observed were quite similar. Thus, only
the microstructures in a keyhole laser weld in alloy 5083 are presented here as .
representative samples of both the alloys. Figure 5.17 shows the base metal
structure of the annealed alloy showing light outlined particles of irregular shaped
Mg 2 Si are seen. Rapid cooling has resulted in the retention of the Mg 2 AI3 phase
in the solid solution. Elongated grains in the rolling direction are evident.
-7
The weld microstructures are shown in figures 5.18 and 5.19. In figure 5.18,
the unmelted region is on the bottom right end of the picture. The weld structure
Not much variation in the dendrite arm spacing is seen from the edge of the weld
to the center. In this non-heat treatable aluminum alloy, the structures in the
p W
~
MmW
MAXIM~~~~~~~~~~~M ~IrU~ ~~W~ k;P. MWBU 7
rX I PO
AM XC V rqUprAIW vATA- N r,, ,"
123
-t,
S S
'5{
*0
. h
rig. 200) .% R
NO
. ,
55,..,
-...
.- . .- - . - _
o
0
124
%
'.L"
'.
* . .
-.
% S
* - -'.2.
Figure 5.19 Weld metal in the middle of the weld (x 200). ; h't
4N
44.
*Sm~.4~w ~ S~*S(~N S S' - ' 5 *N *. , ..
L L
*~~~~-
L..% L%L '.
125
The parent metal microstructure in the 1.27 cm thick plate in alloy 6061.T6
solid solution with fine gray particles of the intermetallic (Fe,Mn,Cr) 3 SiAli 2 along
4 mm thick plate of the same alloy used for the conduction mode welds, similar
The HAZ microstructures in the thick plate 10 kw, 40 ipm on-focus laser weld
are shown in figures 5.21 and 5.22 representing the overaged region and the fusion
due to further precipitation from the solid solution. In figure 5.22, liquation of
the Al-Mg 2 Si eutectic is evident around the ground boundaries. Also, some dark
The weld structures for the same weld are shown in figures 5.23 and 5.24.
of the figure 5.23, and the columnar dendritic structure is clearly evident in the
weld zone. The equiaxed dendritic nature of the structure in the middle of the
weld is shown in figure 5.24. Again, the primary dendrite arm spacing progressively
For the conduction mode welds in the 4 mm thick plates, the structures were .
slightly different. For the 10 kw, 20 ipm conduction mode weld, figure 5.25 and
5.26 show the same region of the fusion boundary at the bottom of the weld at
HAZ. Also agglomeration of the eutectic into large globules is seen in this region.
-lab. ~ -
49S
Figre
.20Bas meal n alo 6-T (x00
A4",
for
ON
Iftw
-%.A
... ....
-A4
... '4, -A
;s4.
Fiue52 Euae enrtcsrutr nth ideo
wel (x 50)
IV
S (4
V- W ,
.. ..-.
"" "- - " V..,--
iI
tt0
C-N
Io
130
Keller's reagent used for the etching has dissolved the Mg 2 Si and has resulted in r
the agglomeration seen as a void. The weld structures are shown in the figures 5.27
and 5.28. The columnar dendritic structure close to the edge of the weld changes
to more equiaxed dendritic as seen in these figures. Compared to the welds in the
1.27 cm thick plates, the primary dendrite arm spacing is much larger. Also, the
amount of the interdendritic eutectic present is higher in the thinner plate welds.
The finite element program THERM described in chapter 4 was used to model
the laser welds in aluminum alloys. The temperature dependent thermal properties
used for the modeling are shown in table 5.7 for each of the four alloys.1,1B2 As the
chemical composition varies only marginally between the alloys 5083 and 5456, and
the experimental results for the weld bead dimensions are also very similar, these
two alloys were not differentiated for the modeling. Alloy 2219 has slightly higher
three alloys. Also, it has the largest solidification range of 100 "K, as compared to
64 'K for the 5xxx alloys and 70 *K for alloy 6061. The temperature dependent
specific heat data was available only for alloy 2219, and it was therefore assumed
constant for the other three alloys. The finite element mesh of 136 elements and
For modeling the aluminum laser welds, the surface heat flux model was used
for the low penetration conduction mode welds, and the internal heat source model
was used for the deep penetration keyhole welds. A gaussian distribution of energy
was assumed in the laser bcam for modeling these welds. The effective beam spot
radius ro for the on-focus and off-focus welds was determined from spot burns on
thick acrylic slabs. The beam spot diameters for the on-focus and off-focus welds
.. ..
* 4~~
. V.V
w %~.
--
131
. _.: .
- r. -... . a *
, - -. .._ s - - u . .. .
C V
,- - ', m. .
.-,.
. - - " JJ
. -- 4/. -
, .
- -, - ..,d'jqV - --.- 'm' ,
-'a-- 2
x 500) .
. , .- -t, -. . .. m
.- -. - *t
-
, .
J}* 4 o v.
4/ .-
IleD o un da ry
L a r e r e n di d
te s t r u c t u r e in e a r e w e l d lb
F ig u r e 5 .2 7
P.~
- -1co
C, co r-C.C'
') -UL.
(0
- 0
C c 0
C.Z
a. ____ __ _ _ _ _ _ _
133
C)E j~
o U7
* ~ , '-
Li% O
<o
(II
z El
00
0.
3c V
- C. V)
L) J
Q)C'4 .
0
S'
.%
134
were estimated to be 0.08 and 0.12 cm respectively. The effective depth yo of the
internal heat source was assumed to be 1.27 cm or the plate thickness. For alloy -
2219, this value resulted in overprediction of the weld penetration, and therefore,
5.6.1 Absorptivity
assumed to ensure that only the laser energy actually absorbed by the material is
used in the heat transfer analysis for each of the welds. The overall welding process
values tu be used for the modeling. As Banas 2 ' has discussed, the overall process
efficiency 77 is defined as S
7
77 = 7a.?7m.
a
Here, r7a is the absorptivity which depends on the material and surface charac-
absorbed laser power. Both the melting efficiency and the absorptivity depend on
the welding mode. At low intensity conduction mode welds, the melting efficiency
is lower than at high intensity levels, as more of the heat is conducted away faster
into the material. Also, the absorptivity changes dramatically with the keyhole
formation as discussed in section 2.3. These two factors result in the overall
process efficiency changing with the welding parameters as seen in figure 5.5. The
absorptivity values used in the calculation of the heat input term as per section %
4.3 were estimated based on the above process efficiency data; absorptivity values
averaging 1.5 times the process efficiency generally gave the best results in the
' ! : t
....-....
""J
b k~"] LI" i, I " '11 I .. . : I" - : : ...- Ai
135 4
Laser welds in aluminum alloys were analysed using the finite element program 4.
THERM, and the predicted weld geometries compared with the experiment-l4
results. The solidus isotherm was defined as the boundary of the weld zone.
test for any weld heat transfer model. As discussed in chapter 4, in the cross- ,
sectional approach for modeling partial to full penetration welds ill thick plates,
the thermal history in the reference plane is predicted at successive time steps as
the heat source moves towards and passes the reference plane. A three dimensional '-
picture of the weld pool can thus be assembled as is shown in the figure 5.30 for a
10 kw, 1.27 cm/sec (30 ipm) on-focus weld in alloy 2219. For clarity, the top view
and the longitudinal section of the same weld pool are shown in figures 5.31 and a-
5.32 respectively. It is seen from figure 5.32 that the maximum weld penetration
in the weld occurs slightly after the heat source has passed the reference plane (0.
in the x-axis). But, the maximum top bead width is seen in figure 5.31 to occur
a considerable distance behind the peak penetration plane. The predicted weld
cross-sections are obtained by the projection of the three dimensional weld pool
Table 5.8 shows the selected aluminum laser welds modeled covering a wide S
range of welding parameters. Also shown in the table are the absorptivity values
used for modeling these laser welds; these values were derived from the process
efficiency data discussed earlier. The type of heat source assumed (surface heat ,._
flux or internal heat source) for each of the welds is also indicated. Figures 5.33
to 5.40 show the weld cross-sections of these welds in the alloys 2219, 5xxx and
6061. The predicted weld profiles match reasonably well with the experimental ]
results. The weld penetration, top bead width as well as the weld cross-sectional %
136
CccN
N4N
-S 0
N_ E 0
u C)
a)
C)
.k%
137*
000
Li-
Lo%
..
...... 37
0S
0 .6LI
I, %
~%
~ lop
I0,
qlp~m ,.a
VC"W
0 2- L
138
's
NN
UJ-I
L~4i)
~ .~* -
0S
0
c
040
Ci
-% ,
LO C)
LA
(ww)U014J40ud
C
(NN
N
coC
0 I0
=1 14
b4ILo
00 00
ol
uojojaa. -wj
140
CNC
E J--
'hi
0~%
1 . C1 0(
(LULU
qua UO 'Dj
141
uK
0 0
IE
coo
S 0
Eu
-- C-
(.w CoI~lqe
0%
142
040
VV
EE
Z0
U.U,
(0
CC)
17 ~z 0 ) .
I CN to co r4
i0
(ww)uoilj~aud LL
143
CN
u
Q)
cc E
PI
ZliQ C)
x
x
- 0
-o C1 0L
Uuw 'all~oadL
%1
144
mS
UJ UU
0 4
LOl
Sx
CD u
co
(fl
Q)%
3b
cc cv I.
ad Vww -a 5'j
V I .,
ILIdo
145
CN
CU)
cU
z m Q
UJ LLJ-
'C
S
C)
(N (Dc
(ujw)uollojqu-v
V%
%01 ld S
I
146
EE
I I.
- Q)
00
00
coo
Di C)UW I. J'
N, .1 .
147
area are in good agreement with the experimental data. However, the model does
(on-focus)
(off-focus)
The finite element program THERM predicts the temperatures at all the
nodes for successive time steps, and from these temperature histories, the thermal,% , ,.
gradients and the local cooling rates are determined. The model was used
alloys. Local solidification times and thereby the solidification rates significantly
influence the weld metal microstructure and the solidification morphology. The
dendrite arm spacings are an important characteristic of the weld metal and
effect its mechanical properties - in general, the finer the spacings , the better
the strength and toughness. The primary dendrite arm spacing is inversely
proportional to square root of the thermal gradient, whereas the secondary arm
%
_ A
148,
experimental data exists for the aluminum - 4.5% copper system which shows
that the secondary dendrite arm spacing and the local solidification time follow S
d =7.50039 (5.1)
where d is the secondary dendrite arm spacing in microns, and 0 the local
Table 5.9 shows the experimentally determined arm spacing at three locations
in a 10 kw, 0.847 cm/sec on-focus laser weld in alloy 2219. Also shown are
the calculated secondary dendrite arm spacings using equation 5.1 and the
solidification times predicted by THERM. The predicted arm spacings are in good
6%
I -d
%
149
Chapter 6
CONCLUSIONS
I
laser welding parameters on the weld characteristics in aluminum alloys 2219,
i. Weld penetration varied considerably among the four aluminum alloys inves- 0
tigated. Alloy 2219 showed the smallest weld penetration and the 5xxx series
alloys the largest for similar welding condition. Presence of volatile alloying
elements like magnesium in the 5xxx alloys alloys results in a higher vapor _1
the laser welds in aluminum alloys exhibited large bead widths owing to the
high thermal diffusivity as well as the lower enthalpy per unit volume. The
overall process efficiency determined from the weld cross-sectional areas also
showed a wide variation among the alloys investigated. The highest process
efficiency of over 33% was observed in 10 kw laser welds in alloy 5456. However,
the process efficiency for alloy 2219 was only about 24% for similar welding
conditions. Conduction mode welds had efficiency values under 5%. The .. ,
areas is directly related to the amount of laser energy absorbed by the material
%. r
ii. The 5xxx series alloys showed a cyclic variation in the weld penetration along S
. . S5* *, -- % *
--. S.-...- -. - . . .- - .. ,C% .-
150
oscillating keyhole. Violent eruptions occurred during the welding cycle due 1
to the frequent collapse of the keyhole and the weld surface often exhibited
open cavities. %
iii. The laser welds in aluminum alloys were characterized for the nature and
amount of porosity present in the welds in terms of the welding parameters and
alloying content. Conduction mode welds were generally free from porosity or
shrinkage cavities. The volume of porosity was the highest for the 5xxx alloys,
and generally decreased with increasing welding speed. In the keyhole welds,
clustered near the top of the welds. Further, laser welds ion these alloys
also contained large cavities in the weld. These cavities are caused by the
entrapment of parts of the unstable cavity as it oscillates between maximum
depth and collapse. Alloys 2219 and 6061 did not exhibit any spherical
iv. Loss of elements from the keyhole laser welds were determined by the electron
microprobe analyzer. Aluminum was the main specie vaporized during laser
welding of alloy 2219. Percentage loss of aluminum from the weld was related
5xxx, both aluminum and magnesium were the dominant vaporizing species.
and the amount of porosity measured could be established. The large amount
of magnesium from the weld metal, and the subsequent entrapment of the
0k
V*.i4* .d'~ r2%
V, % %-~~
bubbles. The large solidification range in the aluminum alloys results in a
Or
large mushy zone during solidification, increasing the probability of bubble
entrapment.
transient heat flow in welding. The model was tested for stability and accuracy -A.-
by solving numerous heat-conduction problems involving phase change. The -
vi. The heat transfer model was used to simulate the melting and solidification
phenomenon occurring in keyhole laser welds. The weld bead shape and di-
mensions predicted by the model are in good agreement with the experimental
results. The model was used to predict the three-dimensional shape of the
weld pool, and the shape of the large mushy zone that results during the
vii. The model was also used to predict the local solidification rates and times V",
zone were correlated with the secondary dendrite arm spacings in the weld
, % . ..
%.R I
J.
-IF _V
-v K-. -i
152
REFERENCES
3. Breinan, E.M., Banas, C.M., and Greenfield, M.A. 1981. Laser Welding
- The Present State-of-the-Art. Source Book of Electron Beam & Laser Welding,
pp. 247, ASM, Metals Park, OH. 6
9. Snow, D.B., and Breinan, E.M. 1978. Evaluation of Basic Laser Welding
Capabilities. United Tech. Research Center Report R78-911989-14.
10. Snow, D.B., Kaufman, M.J., Banas, C.M., and Breinan, E.M. 1979. -
Evaluation of Basic Laser Welding Capabilities. United Tech. Research Center
Report R79-911989-17.
11. Moon, D.W., and Metzbower, E.A. 1983. Laser Beam Welding of
Aluminum Alloy 5456. Welding Journal,62(2):53-s to 58-s.
12. Oreper, G.M. and Szekely, J. 1984. Heat- and Fluid-flow Phenomena in
Weld Pools. J. Fluid. Mech, 147:53-79.
13. Marcus, S., Lowder, J.E., and Mooney, D.L. 1976. Large Spot Thermal
Coupling of CO 2 Laser Radiation to Metallic Surfaces. J. Appl. Phys., 47(7):2966-
2968.
JS W
153
14. Pavelic, V., Tanbakuchi, R., Uyehara, O.A., and Myers, P.S. 1969. **1
15. Tong, H., and Giedt, W.H. 1971. Depth of Penetration during Electron
Beam Welding. J. Heat Transfer, 93(5):155-163.
16. Miyazaki, T., and Giedt, W.H. 1982. Heat Transfer from an Elliptical
Cylinder Moving Through an Infinite Plate Applied to Electron Beam Welding.
Int. J. Heat Mass Transfer, 25(6):807-814.
17. Mazumder, J., and Steen, W.M. 1980. Heat Transfer Model for CW Laser
Material Processing. J. Appi. Phys., 51(2):941-947.
19. Goldak, J., Chakravarti, A., and Bibby, M. 1984. A New Finite Element
Model for Welding Heat Sources. Metall. Trans. B, 15B(6):299-305. 0
20. Goldak, J., Bibby, M., Moore, J., House, R., and Patel, B. 1986. Computer
Modeling of Heat Flow in Welds. Metall. Trans. B, 17B(9):587-600.
21. Ready, J.F. 1982. Material Processing - An Overview. Proc. of the IEEE, 0
70(6):533-544. i_
22. Huntington, C.A., and Eager, T.W. 1983. Laser Welding of Aluminum
and Aluminum Alloys. Welding Journal, 62(4):105-s to 107-s.
26. Lewis, G.K., and Dixon, R.D. 1985. Plasma Monitoring of Laser Beam
Welds, Welding Journal, 64(2):49-s to 54-s.
27. Dixon, R.D., and Lewis, G.K. 1985. Electron Emission and Plasma "'na 4
Formation during Laser Beam Welding. Welding Journal, 64(3):71-s to 78-s.
28. Martukanitz, R.P. 1982. Sources of Porosity in Gas Metal Arc Welding of
Aluminum. Trends in Welding Research, pp.315-330, ASM.
154
29. Andrew, R.C., and Chadwick, G.A. 1975. Source of Hydrogen Porosity
in Aluminum Alloy Welds - A Literature Survey - AWRA Report P3-5-74.
X,
Australian Welding Research, pp. 1-19.
30. Talbot, D.E.J. 1975. Effects of Hydrogen in Aluminum. Magnesium,
Copper, and Their Alloys. Int. Metall. Reviews, Review 201, 20:166-184.
31. Hatch, J.E. 1984. Aluminum, Properties and Physical Metallurgy, ASM.
32. Howden, D.G., and Milner, D.R. 1963. Hydrogen Absorption in Arc
Welding. Brit. Weld. J., 10(6):304-316.
36. Uda, M., and Ohno, S. 1974. Porosity Formation in Weld Metal - I.
Effect of Hydrogen on Porosity Formation in Pure Aluminum At Non-Arc Melting.
Trans. National Res. Inst. for Metals, 16(2):29-36.
37. Nikiforov, G.D., Trusov, S.A., and Silant'eva, S.A. 1980. The Redistribu- 0
tion of Hydrogen during the Solidification of Aluminum. Svar. Proiz., 9:44-48.
39. Opie, W.R., and Grant, N.J. 1950. Hydrogen Solubility in Aluminum and
some Aluminum Alloys. Trans. of AIME, J. of Metals, 188(10):1237-1241.
41. Chalmers, B. 1964. Principles of Solidification. John Wiley & Sons, NY.
42. Kubo, K., and Pehlke, R.D. 1985. Mathematical Modeling of Porosity.*-"...:,
Formation in Solidification. Metall. Trans. B, 168(6):359-366.
44. Mazumder, J. 1982. Laser Welding: State of the Art Review. J. Met.,
34(7):16-24.
155
45. Block-Bolten, A., and Eager, T.W. 1982. Selective Evaporation of Metals
from Weld Pools. Trends in Welding Research, , pp.53-73, ASM.
0
46. Khan, P.A.A., and DebRoy, T. 1984. Alloying Element Vaorization and
Weld Pool Temperature during Laser Welding of AISI 202 Stainless Steel. Metall.
Trans. B, 15B(12):641-644.
47. Schauer, D.A., Giedt, W.H., and Shintaku, S.M. 1978. Electron Beam
Welding Cavity Temperature Distributions in Pure Metals and Alloys. Welding
Journal, 57(5):127-s to 133-s..60111
50. Myers, P.S., Uyehara, O.A., and Borman, G.L. 1967. Fundamentals of
Heat Flow in Welding. WRC Bulletin 123, Welding Research Council, NY.
52. Malmuth, N.D. 1976. Temperature Field of a Moving Point Source with
Change of State. Int. J. Heat Mass Transfer, 19(4):349-354.
53. Trivedi, R., and Srinivasan, S.R. 1974. Temperature Distribution Around
a Moving Cylindrical Source. J. Heat Transfer, 96(8):427-428.
54. Swift-Hook, D.T., and Gick, E.F. 1973. Penetration Welding with Lasers.
Welding Journal, 52(11):492-s to 499-s.
55. Gray, W.H., Schnurr, N.M. 1975. A Comparison of the Finite Element
and Finite Difference Methods for the Analysis of Steady Two Dimensional Heat 0
Conduction Problems. Comp. Methods in Appl. M[ech. and Engn., 6:243-245.
57. Zienkiewicz, O.C., and Parekh, C.J. 1970. Transient Field Problems: Two- "-
" "' "
Dimensional Analysis by Isoparametric Finite Elements. Int. J. of Num. Methods
in EngN., 2:61-71. 0
58. Hsiao, J.S. 1985. An Efficient Algorithm for Finite-Difference Analysis of
Heat Transfer with Melting and Solidification. Num. Heat Transfer, 8:653-666.
S
156
59. Friedman, E. 1974. A Direct Iteration Method for the Incorporation of
Phase Change in Finite Element Heat Conduction Programs. AEC Research and
Development Report WAPD-TM-1133. S
60. Comini, G., Del Guidice, S., Lewis, R.W., and Zienkiewicz, O.C. 1974.
Finite Element Solution of Non-Linear Heat Conduction Problems with Special
Reference to Phase Change. Int. J. for Num. Methods in Engn., 8:613-624.
61. Pham, Q.T. 1987. A Note on Some Finite-Difference Methods for Heat
Conduction with Phase Change. Num. Heat Transfer, 11:353-359.
64. Wood, W.L., and Lewis, R.W. 1975. A Comparison of Time Marching
Schemes for the Transient Heat Conduction Equation. Int. J. for Num. Methods
in Engn., 9:679-689.
66. Hultgren, R., Orr, R.L., Anderson, P.D., and Kelley, K.K. 1963. Selected
Values of Thermodynamic Properties of Metals and Alloys. John Wiley and Sons, r
NY.
67. Dushman, S., and Lafferty, J.M. 1962. Scientific Foundations of Vacuum
Technique. 2nd ed., John Wiley and Sons, NY.
68. Cobine, J.D., and Burger, E.E., 1955. Analysis of Electrode Phenomenon
in the High-Current Arc. J. Appl. Phys., 26(7):895-900.
71. Rao, S.S. 1982. The Finite Element Method in Engineering. Pergamon
Press, NY.
157
72. Wilson, E.L., Bathe, K.J., and Peterson, F.E. 1974. Finite Element
Analysis of Linear and Nonlinear Heat Transfer. Nuclear Engn. and Design,
29:110-124. 6
74. Weiner, J.H. 1955. Transient Heat Conduction in Multiphase Media. Brit.
J. Appl. Phys., 6(10):361-363.
78. Krutz, G.W., and Segerlind, L.J. 1978. Finite Element Analysis of Welded
Structures. Welding Journal,57(7):211s to 216s.
79. Chong, L.M. 1982. Predicting Weld Hardness. M. Eng. Thesis, Carleton
Univ, Ottawa, Canada.
80. Locke, E., Hoag, E., and Hella, R. 1972. Deep Penetration Welding with
High Power CO 2 Lasers. Welding Journal, 51(5):245-s to 249-s.
81. Touloukian, Y.S., and Ho, C.Y. 1973. Properties of Aluminum and
Aluminum Alloys. Thermomechanical Properties Research Center, Purdue Univ.,
IN.
83. Kurz, W., and Fisher, D.J. 1986. Fundamentals of Solidification. Trans
Tech Publications, Switzerland.
.% *
V - J~'<~
~ V ~V ~ V ~ %.
158
APPENDIX
input data sets are required. The fi-st called INPUT consists of a description
of the overall problem and domain, the welding parameters including the terms
describing the spatial distribution of the energy, and the material thermal
properties. The second data set called MESH describes the finite element mesh S
used to discretise the domain being analysed. Following sections briefly describe
LNPUT
A sample INPUT file for a 10 kw, 1.27 cm/sec (30 ipm), on-focus laser weld
Each line in the above example is described in the following paragraphs and .
I A
tables.
159
1 NELEM # of elements
2 NNODE #of nodes
5 EFF Absorptivity
-' - . q.:
160
Line 4 (Format 5F10.3)
r
Line6 (Format 5F10.3)
1 TS Solidus temperature
2 TL Liquidus temperature "-.'- "
" S
,2'-
eA
161
Line 8 Onwards
Next few lines declare the thermal conductivity, volumetric specific heat and
the enthalpy at N(1), N(2) and N(3) temperatures respectively. Also, line 7
through end are repeated NMATL times for each material in the domain.
MESH
An example of the MESH file used for the finite element mesh shown in figure S
5.33 is listed at the end of this section. The first set of lines (Format 1515) ON
describe the nodal connectivity array for each of the elements. Item 1 is the -
element number. Wherever similar patterns of elements exist, node numbers and
other characteristics of the element are generated. For instance, based on the
values in lines 5 and 6 for elements 5 and 37 respectively, the node numbers etc.
for elements 5, 9, 13, 17, ..... and 37 are generated - item 2 defines the increment
Item 3 in these lines defines NE for each element which identifies any edge in
the quadrilateral element which lies on the surface. This is required for calculating
the convective and radiation losses from the element surface.
S
162
NE Description
I Edge connecting nodes 1 and 2S
2 Edge connecting nodes 2 and 3
order to use the associated material properties. Item 5 gives the number of nodes
(4 to 8) being used in the particular element, and item 6 to last give the associated
In the second set of lines, the nodal coordinates (x and y) are listed (Format
occurs in the mesh using NDIFF - the node numbers increment in the pattern.
1 0 2 1 4 19 1 2 20
2 0 2 1 4 38 19 20 39 "V'
3 0 2 1 4 58 38 39 59
4 0 2 1 4 78 58 59 79 %
5 0 0 1 4 20 2 3 21
37 4 0 1 4 28 10 11 29
6 0 0 1 4 39 20 21 40
34 4 0 1 4 46 27 23 47
7 0 0 1 4 59 39 40 60
31 4 0 1 4 65 45 46 66
8 0 0 1 4 79 59 60 80
28 4 0 1 4 84 64 65 85
32 0 0 1 4 85 65 66 67 0
35 0 0 1 4 66 46 47 48
36 0 0 1 4 67 66 48 68
38 0 0 1 4 47 28 29 30
39 0 0 1 4 48 47 30 49
40 0 0 1 4 68 48 49 50
41 0 0 1 5 30 11 12 31 29
42 0 0 1 5 50 30 31 51 49
43 0 0 1 4 31 12 13 32
53 2 0 1 4 36 17 18 37
44 0 0 1 4 51 31 32 52
54 2 0 1 4 56 36 37 57 %
55 0 1 1 5 78 80 87 86 79
56 0 0 1 5 80 82 88 87 81
57 0 0 1 4 88 82 83 89 0
59 1 0 1 4 90 84 85 91
60 0 0 1 4 91 85 67 92
',% %
bOUNWAMI U M~J M j7777.p-7'.77777 -7
77 -if.~1.VKJ v v -. ,Te -
163
70
71
0 :
30 - 4
4
9
,7
-5-7
-J
94
9j
.
_
2 70
9
763 0 C 4 9 '
85- 0 0 1 5 9. S7 e q9 9
70 0 0 2. 4 99 92 69 %00
77 0 0 1 4 2.00 69 70 101
85 0 1 1 5 93 95- 110 109 94
86 0 0 1 5 95 7 11 110 96
87 0 0 1 5 97 99 112 112 98
89 0 0 ! 5 99 112 112 10000
89 0 0 1 5 10 13 113 102
90 0 0 4 214 103 1-04 1-1-5
94 1 4 11 107 0 119
95 0 2 1 4 120 109 110 121
96 0 0 . 4 1 1 -
1 0.00000 0.6950
13 1 0.00000 0.00000
i9 0 0.02000 ,.27000
23 1 .02000 1.14300
0 0.02000 1.07953
29 1 0.02000 0.76200
30 0 04000 0.76200
1_ 0.04000 0.63500
33 0 0.04000 0.50800
37 0.04000 0.00000
39 0 0.04000 1.27000
42 i 0.04000 1.14300
43 0 0.04000 1.079503 .
47 - 0.04000 0.825 'o
48 0 0.06000 0.8:5-0z
49 0 0.0600 0.76200
50 0 0.08000 0.76200
em
rv ~ ~ r~ ~ 4~ 'q.
164
2 0.08000 06 3 500a
--
0 0.08000 . 3060C
'....,.)-.
570.08000 0.0000
53 0 0.06000 Z.7000
62 0.0000 . 00
3 0 0.06000 .50
66 2 0.06000 0.88900
67 0 0.08000 0.8900
63 0 0.08000 0.8250
69 0 0 .2000 0.82550
72 2 0.2000 0.63500
73 0 0.82.000 0 .800
77 2 0.2.2000 0 . 00000
78 0 0.08000 1.27000
82 1 0.08000 1.24300
83 0 0.08000 1.07950
85 . O0.08000 0.95250
86 0 0.2000 1.27000
92 2 0.22000 0.8800"
93 0 0.6000 1.27000
503 1 0.16000 0.63500
2.04 0 0.126000 0.50800
.08 1. 0.126000 0.00000
209 0 0.28000 1.27000
"1 0.8 000 O.00000
2.0 0 0.40000 1.27000
10 1 0.40000 0.00000
132. 0 0.60000 1.27000
136 1 0.60000 0.00000
137 0 0.30000 1.27000
142 0.80000 0.00000
143 0 1.00000 1.27000
148 1 1.0000 0.00000
.49 0 2.00000 1.27000
254 1 2.00000 0.00000
2.5 0 4.00000 1.27000
167 4 1.000000 .27000
.56 0 4.00000 1.01600
53 4 2.00000C 701600
!57 0 4.0000C 0.50800 , "
-69 4 !0.00000 0.50800 ."" . "
153 0 4.00000 0.00000
170 4 10.00000 0.00000
,,OUTP UT 0 U,,
,.,
,, _.
,0,',